Winning on

HR
Analytics
Leveraging Data for
Competitive Advantage

Ramesh Soundararajan
Kuldeep Singh
SAGE was founded in 1965 by Sara Miller McCune to support
the dissemination of usable knowledge by publishing innovative
and high-quality research and teaching content. Today, we
publish over 900 journals, including those of more than 400
learned societies, more than 800 new books per year, and a
growing range of library products including archives, data, case
studies, reports, and video. SAGE remains majority-owned by
our founder, and after Sara’s lifetime will become owned by
a charitable trust that secures our continued independence.

Los Angeles | London | New Delhi | Singapore | Washington DC | Melbourne

 Advance Praise

This book provides HR Analytics techniques and very practical

set of action oriented recommendations to leverage human talent.
Srinivas Kandula,
CEO, Capgemini, India

Ramesh and Kuldeep have filled this book with helpful and timely
examples of leveraging analytics in Human Resources today.
Analysts, benefit from their research and help your organization
further its goals.
Jeremy Shapiro, Executive Director,
HR, Morgan Stanley

This book provides broad insights to this emerging field and prac-
tical guidance and advice for every HR practitioner.
Marc Effron, President,
The Talent Strategy Group, New York

HR is one of the fastest-growing areas for analytics, and this is

an invaluable guide to the subject. If you want to hire, retain, and
motivate the best people, you need to read this book and follow
its advice.
Thomas H. Davenport, Distinguished Professor,
Babson College, Author of Competing on
Analytics and No Humans Need Apply
In a rapidly moving and advanced field like HR Analytics, there
is always a need for new and useful up-to-date content and learn-
ing. This book adequately and provocatively fills this space bring-
ing new perspectives and practical ideas for HR and analytics
professionals.
Max Blumberg, PhD, Analytics Advisor to the CIPD,
Management Consultant, and Visiting Researcher,
Goldsmiths, University of London

We often miss the strategic and financial value of insights into

our organization’s workforce. This book provides a framework
for extracting and putting them to use.
John Cunnell, Serial Entrepreneur
Winning on

HR
Analytics
Copyright © Ramesh Soundararajan and Kuldeep Singh, 2017

All rights reserved. No part of this book may be reproduced or utilized in any
form or by any means, electronic or mechanical, including photocopying,
recording, or by any information storage or retrieval system, without permission
in writing from the publisher.

First published in 2017 by

SAGE Publications India Pvt Ltd

B1/I-1 Mohan Cooperative Industrial Area
Mathura Road, New Delhi 110 044, India
www.sagepub.in
SAGE Publications Inc
2455 Teller Road
Thousand Oaks, California 91320, USA
SAGE Publications Ltd
1 Oliver’s Yard, 55 City Road
London EC1Y 1SP, United Kingdom
SAGE Publications Asia-Pacific Pte Ltd
3 Church Street
#10-04 Samsung Hub
Singapore 049483

Published by Vivek Mehra for SAGE Publications India Pvt Ltd, typeset in
11/13 pt Times New Roman by, Fidus Design Pvt. Ltd., Chandigarh 31D and
printed at Saurabh Printers Pvt Ltd, Greater Noida.

Library of Congress Cataloging-in-Publication Data Available

ISBN: 978-93-860-4241-5 (PB)

Sage Team: Sachin Sharma, Priya Arora, Megha Dabral and Ritu Chopra
Dedicated to objectivity and transparency
in people management
 Contents

Foreword by Alec Levenson xi

Preface xvii
Acknowledgments xxi

1. It Is the Right Time for Analytics in HR 1

HR’s Tryst with Competitive Advantage 1
Human Capital Alone is Not Sufficient 2
HR Policies Are Critical Too 3
What Is HR or People Analytics? 6
Why This Sudden Interest in HR Analytics? 7
Big Data Era and HR Analytics 10
Business Strategy–HR Analytics–Competitive Advantage
Integration 12
2. Articulating Business Value of HR Programs 14
HR Analytics Linkage to Business Outcomes 16
Measuring Use of HR Analytics Impact on Business
Outcomes 16
Measuring HR Programs for Business Results Linkages 18
Research Evidence on Impact of HR Programs 23
How to Measure Linkage of HR Programs to Business
Outcomes? 24
Industry Examples of Measuring HR Programs Impact 29
3. Analytical Problem Solving 31
Deep and Wide Approach 31
Building the Cube 37
viii Winning on HR Analytics

4. Competing Through Workforce Analytics 47

Business Levers of Organization Structure 47
Traditional Measures of Organization Structure 48
Becoming More Competitive Using Organization Structure 51
Organization Shaping and Employee Growth 56
Look at Headcount in Offices 59
Measuring the Softer Aspects of Organization Structure 60
Organization Demographics and Succession Planning 60
5. Acquiring High-quality Talent 64
Business Levers of Talent Acquisition 64
Traditional Measures of Talent Acquisition 65
Effectiveness Measures 68
Emerging Measures of Talent Acquisition 71
Opportunity Cost of Cycle Time 72
Validity of Hiring Specifications 73
Importance of Quality of Hire 75
Talent Acquisition for Predictable Joining and Performance 78
Measuring and Improving Process Capability 81
6. Results-oriented Talent Development 84
Measuring Return on Investments on Talent Development
Initiatives91
Right Metrics and Measures for Strategic Alignment 93
7. Talent Engagement and Retention 98
Business Levers of Employee Engagement 98
Traditional Measures of Engagement 102
Measuring Attrition 102
LTM or YTD? 104
Employee Retention 106
Predictive Modeling for Attrition Analysis 121
8. Measuring and Managing Competencies 124
Competency Baselining 125
Usage of Competency Baselines 128
Leadership Development 130
Using Competencies in Talent Acquisition  131
 Contents ix

9. Optimizing Compensation and Benefits for

High Performance 133
Business Levers of Compensation and Benefits 134
Organization Structure and Cost of Management 135
Traditional Measures of Compensation 139
How Far Does Annual Compensation Increase Help? 140
We Are a High Performance Organization. Are You Sure? 144
Valuing Benefits Using the CTC Statement 147
Portfolio Management of Benefits 148
Tailoring Variable Pay to Performance Based on Data 151
10. Making the Transformation Possible 152
Executing Transformation—Rubber Hits the Road 162
People Analytics: Hype Versus Truth 164
Appendices169
Appendix A: How to Get Started in HR Analytics 172
Appendix B: Seven Deadly Sins of HR Analytics
Initiatives 177
Appendix C: Starting with Workforce Analytics?
Five Considerations Before Taking the Leap 183
Appendix D: Small Data can be as Powerful
as Big Data186
Appendix E: Establishing RoI for Training Investments191
Appendix F: Why Perception is Important for People
Analytics197
Appendix G: Case Study for Talent Acquisition 202
Appendix H: Case Study for Building a Business
Case for Employee Retention 208
Appendix I: Using Statistics to Arrive at Engagement
Drivers 213
x Winning on HR Analytics

Appendix J: Making the Case for Predictive Attrition

Risk Modeling: A Roadmap for the Future 219
Appendix K: Discovering Team Cohesiveness and
Influencers Using Organization Network Analysis 230
References236
Index240
About the Authors244
 Foreword

I t has been a half century since HR was known as the personnel

function, and two decades since Dave Ulrich challenged HR
to get a seat at the table. As part of the evolution of the func-
tion toward being more strategic, we have moved away from
an emphasis on basic measurement to scorecards, engagement
surveys, and strategic workforce planning. Today, these activities
are all grouped under the umbrella of HR analytics.
Despite the enormous attention being paid to HR analytics
today, there is a good deal of confusion regarding where people
should be focusing their attention and what they should be doing.
As Soundararajan and Singh note in the Preface, a lot of what
exists in HR today can be traced back to scientific research that
occurred at some point in the past. And what is not explicitly
based on research usually has a strong measurement component.
Data and analysis have been a part of HR for as long as the func-
tion has existed. So what is new about the current emphasis on
HR analytics?
I see the current excitement and energy arising from converg-
ing trends in strategic HR, computing/technology innovation, and
an appreciation for the benefits of learning from proven practice
(evidence-based HR).
On the strategic front, HR has been searching for the longest
time for the secret sauce that will enable it to be more strategic.
There has been quite a bit of progress, but at the same time, there
have been a lot of frustrations as well. The survey of the state of
the HR function conducted by my colleagues, Ed Lawler and John
Boudreau, at the Center for Effective Organizations over the past
20 years has shown surprising little change in the amount of time
people in HR spend on strategic versus transactional activities.
xii Winning on HR Analytics

It’s possible to read this as a lack of progress in becoming more

strategic, but I have a different take on this.
There is a lot of basic work that has always been and will always
be part of the work of HR, most of which does not seem parti-
cularly strategic at first glance: making sure that people are paid
properly, open positions are filled, performance reviews are con-
ducted, development planning takes place, and much more. Under
certain circumstances, these activities can be strategic, yet most
of the time they are more about “keeping the lights on”—enabling
the business to do its work by ensuring that people are in place to
do the work when and where it needs to happen. Sometimes, when
there is a critical business need best served by these traditional
HR practices, doing this everyday work of HR is strategic. So,
whether traditional HR is truly strategic or not often depends on
the context. One job for HR analytics is to understanding when
and where that is the case.
The second trend is the rapid development and deployment
of technology that makes it easier to collect and warehouse data
in easily accessible formats. This includes both the proliferation
of survey vendors and do-it-yourself Internet-based survey tools.
It also includes the widespread installation of enterprise resource
planning (ERP) and other business IT systems that link together
for joint analysis of previously disparate data systems that were
hard to integrate.
On the employee side, the now widespread ability to survey
people has easily led to an explosion of surveys conducted both
internally and by outside consultants. It seems that everyone wants
to measure as much as possible related to people, in the hopes
that something will emerge that will be useful. Yet the often-cited
problem of survey fatigue is a telling sign that we have too much
measurement that is not being guided by the right questions and
models.
On the business IT side, there has been an enormous shift-
ing of priorities for many HR functions, with the cart too often
being placed before the horse. The promise of the ERP systems,
along with their outrageous price tags, creates a set of “facts on
the ground”: warehoused data that is expected to be analyzed first
and foremost before turning to other data sources. This happens
 Foreword xiii

for two reasons. One, because the data is readily available, it is

very tempting to dive right into mining it for interesting patterns,
a temptation that most data scientists know can be very hard to
resist. Second, the obscene sums spent in installing the systems
create enormous pressure on the HR function to do something
with the data to justify at least part of the sunk costs, which usually
were authorized outside the HR function in the first place. Rather
than question the wisdom of focusing on that data, HR dutifully
falls in line and dives right into mining it for interesting patterns,
even when there is no strategic compass to guide the work.
The third trend is the increased awareness of the importance of
practicing evidence-based HR. In truth, this trend is more aspi-
rational than widespread, getting more attention in the academic
and research communities than within the HR function itself. Yet
the growing number of data scientists and people working in HR
with advanced degrees in industrial-organizational psychology
and other fields has provided a good deal of internal momentum
toward taking a more scientifically valid approach to defining and
analyzing HR issues.
The good news is that there’s a lot of evidence to draw upon
to improve management practices such as goal setting, allocat-
ing rewards, doing employee selection, allocating training invest-
ments and more. Yet the information on the evidence is usually
not communicated in ways that make it widely accessible to a
broad management audience, and, even worse, little to no guid-
ance is provided on how to prioritize what HR and the business
should focus on. Consequently, the messages that emerge from
the scientific community about how analytics can improve HR
and management practice are disjoint and not focused directly on
pressing business issues.
Even worse, many of the data scientists and social scientists
with advanced degrees who work in and consult with organizations
do not take enough of a systems perspective when approaching
the analysis of HR issues. They too often settle for incrementally
better (more scientifically valid) measurement approaches with-
out first ensuring that the most important, pressing business issues
are being addressed.
xiv Winning on HR Analytics

Today there are tons of data available that measure the execu-
tion of HR processes: headcount, vacancies, time to fill, comple-
tion of performance reviews, distribution of performance ratings,
details on individual development plans, and so on. The problem
with the current practice is that HR analytics is used to describe
these processes without embedding the inquiry in a strategic con-
text. This means that the analysis often reveals data patterns that
can seem interesting but more often than not elicit a “so what”
response: What is the value in looking at the data? Where are the
insights that can help the business to function more effectively?
To address these questions and ensure that HR analytics adds
maximum value, there are three steps to follow: (a) ask the right
questions, (b) do the right analysis, and (c) lead the change. Of
these three, only the second is done today in HR analytics with
any regularity, but even then common practice falls short of the
ideal. This book and other contributions make important advances
in this area, but with some critical caveats because common prac-
tice is not changing fast enough. On the other two fronts, there
has been very little progress except in rare instances, with the
exceptions proving the rule.
Start with asking the right questions. For me, the most impor-
tant place to start for any HR analytics inquiry is the hypotheses
being tested. What is the main purpose in doing the analysis?
What business problems are you trying to solve? Are you trying
to improve the current HR practice to make it more efficient and
effective? Are you trying to help the business to improve strategy
execution?
Asking the right questions often requires looking beyond the
specific request that is made regarding HR analytics to get at
what’s really at the heart of the matter. For example, “how do we
improve employee engagement” at face value can sound like “how
do we improve employee morale” or “how do we get our people
more actively involved in providing discretionary effort?” Faced
with that request, most HR analytics practitioners will charge ahead
and look only at how people feel about the work they are doing
and search for ways to improve their attitudes and motivation.
Such a pursuit is worthwhile—if indeed employee engagement is
 Foreword xv

the ultimate end result that the business needs. Yet in most cases,
engagement is not the end result but instead, one of the contribu-
tors to performance, and it is performance that is the real target.
As detailed in my book Strategic Analytics, to answer such a ques-
tion, you need to take a more systematic look at the factors driv-
ing performance at the individual level, and broaden the scope of
the HR analytics inquiry to include the work design and the
competencies of the people in the role.
Soundararajan and Singh set the stage the right way by putting
the discussion of how to link HR analytics to business outcomes
at the beginning of this book. What the reader should know as you
dive into the content is that there are multiple ways to frame and
address business impact. Whether it’s the approach taken in this
book, in my book, or any of a number of other ways, choose the
one that works for you and makes the most sense to your stake-
holders and business partners. It’s the destination that matters
more than the path chosen to get there.
When it comes to doing the right analysis, there are more dif-
ferent types of analytics that can be conducted. Trying to sort
through them all is very daunting if you start from the perspective
that you need to have a good grasp of all the different types of
ways HR analytics has been applied—and especially if you feel
like you need an advanced degree in statistics to make sense of
it all. My advice here is (a) stick closely to the questions at the
core of your inquiry, (b) find the right data to answer them, and
(c) don’t choose elegance of the analytical method over a laser-
like focus on answering the questions. For example, many of the
analyses presented in this book are very simple, consisting of
calculating ratios or showing data patterns in a table or graph. If
you are asking the right question and have the right data to answer
it, those types of analysis are often all that you need to do. And
to that toolkit I would add diagnostic interviews which often are
the only way of analyzing issues like organization design and
alignment, cross-functional effectiveness, and strategy execution
at the business unit level.
The last key for doing HR analytics the right way is integrat-
ing the analysis with organizational change processes. To be most
xvi Winning on HR Analytics

effective, this means starting with the end in mind: no HR analytics

analysis should ever be undertaken without a clear understanding
of how the analysis will be used, including knowing how the rele-
vant stakeholders will react when presented with the information.
This means that the business case for doing the analysis needs to
be already known ahead of time, or needs to be established jointly
with the relevant stakeholders. This last foundation for doing HR
analytics the right way is usually the one least followed, leading
many, many analyses to fall on deaf ears: they can generate some
interest but often little to no action that makes a difference. If this
sounds to you like I am promoting good old-fashioned organiza-
tion development (OD) and change management, you are correct:
the most effective HR analytics processes incorporate those core
OD principles.
The journey to more effective HR analytics will not be com-
pleted in a day, month or even year. And the tools and resources
needed cannot ever be contained in one volume. This book can be
a very useful contributor as you make your way on that journey,
so long as you keep in mind the big picture of what you’re trying
to accomplish, how you can best serve the organization’s larger
strategic needs, and how your work in HR analytics fits in.

Alec Levenson,
author of Strategic Analytics:
Advancing Strategy Execution and
Organizational Effectiveness,
Senior Research Scientist,
Center for Effective Organizations,
Marshall School of Business,
University of Southern California,
Los Angeles, CA, USA
 Preface

You see things; and you say “Why?” But I dream things
that never were; and I say “Why not?”
—George Bernard Shaw

H R analytics is in the hype cycle today. There are conferences

around this emerging field. Cool new technology is evolving
that can help organizations visualize their existing data into spark-
ling charts. There seem to be only two kinds of companies: Ones
that use predictive analytics in HR and the ones that are planning
to! Some gurus even hope that HR analytics is the latest tool that
can take the function to the next level.
Are analytics really that new in HR? Unlike many other func-
tions, HR is based on behavioral research. Right from the pioneer-
ing Hawthorne experiments to Theory Y, solid behavioral research
underpins HR challenges such as training and motivation.
Geert Hofstede had set up a personnel research department in
IBM around 50 years back. He arrived at the cultural dimensions
theory based on more than 100,000 surveys. Jac Fitz-enz initi-
ated his outstanding work on HR metrics and measurement in the
1970s. Gallup’s Q12 was based on researching millions of survey
responses in the 1990s. HR scorecard by Brian Becker and Dave
Ulrich was published in 2001.
Taken that way, HR analytics is more an evolution than a revo-
lution. If it is an evolution, what is the need for a book at this point
in time?
When observed with intent, two patterns emerge in HR.
First is the rush to adapt best practices without enquiry. Around
the turn of the millennium, General Electric (GE) emerged as
the benchmark for HR practices. Irrespective of the maturity of
xviii Winning on HR Analytics

business, companies started setting up leadership development

programs. Competencies were identified, leaders were assessed,
and development plans put in place. Yet, even after all these years,
one cannot correlate with certainty whether having a dedicated
leadership development program produces better leaders.
The adaption of Bell Curve is a classic illustration. Every com-
pany had a performance appraisal process and a compensation
review process. Based on their culture and business needs, com-
panies had different levels of interconnect between the two. Many
had a public appraisal rating and a more secretive compensation
decision. Just then everyone read about the great impact of nor-
malization on GE’s performance. HR heads cheered on by CEOs
embraced normalization without really taking a deep breath and
exploring the intended business results. In a sense, if you are not
normalizing, you are not one of us!
Let us flash forward to 2016. While it could have helped GE
with streamlining its workforce, the benefits of using the normal
curve have not been equally visible across the board. Murmurs
had started 7–8 years back on the negative impact of normali-
zation on employee morale. However, companies kept going
with normalization till one of them pulled the plug and announced
that it is not working for them. Adobe, Microsoft, and Deloitte are
some of the high-profile trendsetters. Suddenly traffic is jammed
with companies moving away from normalization!
These are just two examples of how HR organizations have
been adapting practices not based on their own insights, but due
to a bandwagon effect!
This brings us back to HR analytics. Unlike other practices,
analytics is not about identifying a few people, buying some tech-
nology, and making some presentations, though it involves all the
three. This book does not follow the path that you are being left
behind every day if you are not using predictive analytics.
In our personal experience at work, we have been fascinated to
see that:

• College, percentage marks, and performance in aptitude test

have no correlation with on-the-job performance of gradu-
ate engineers, but performance during training has.
 Preface xix

• There is no correlation between percentage salary increase

and retention.
• Employees undergo training programs and their compe-
tency scores actually decline!
• Pride of association with a successful company has a strong
correlation with employee satisfaction.
• There is a correlation between employee satisfaction and
customer satisfaction for a given business group.

Most of this analysis was carried out using the advanced functions
of spreadsheets and simple presentations. Since then, the analyti-
cal and presentation capabilities have improved manifold. At the
same time, it is not a surprise to see even large and successful
companies struggle for reliable data on which they can form their
hypothesis.
Hypothesis is the operating word here. The classic PDCA cycle
emphasizes plan, do, check, and act. You set goals, plan strategies
to achieve them, measure outcomes, and take corrective actions
where required. Analytics can help ask the right questions and
align all the four.
This book is based on our experiences and insights gained from
a cumulative experience of 50 years. It is our conviction that com-
panies can win with analytics. However, that needs a structured
approach based on:

• Planning HR strategy around hypothesis,

• Setting up goals for the strategy implementation,
• Review using metrics,
• Make course corrections based on what metrics say.

For ease, the book is organized into three parts:

• Evolution of HR analytics and establishing the business

case for HR programs using analytics.
• Focus on each talent management process: acquisition,
development, engagement, performance management, etc.
• Summarize with an implementation strategy.
xx Winning on HR Analytics

Some of the processes and implementation are supported with

insights and case studies.
This book should be handy if you are starting off your career
and would like to get a perspective on taking an analytic view
to HR. It would be handy if you are heading an HR function
and would like to improve your performance. Even if you have
a sophisticated analytics operation, we hope you can find some
insights that are relevant.
Again, this is not about the latest and greatest things happening
in the world of analytics. While we have expanded the scope to
include subjects such as network analysis, contextual search, and
text-based analytics, there could be better resources if your inter-
est is solely in leading edge work. However, this is more around
developing an analytical view of the function that leads to an
effective use of what is out there.
Just a question in closing: We had mentioned that authors have
a cumulative experience of 50 years. What exactly does it indi-
cate? Is it better than 40 years’ experience? Would someone with
60 years be better? Or it is five times as valuable as 10 years’
experience?
If you have been asking such questions, we are sure you would
find this relevant! To go back to the famous quote at the begin-
ning, whether “Why” or “Why not,” curiosity to question is where
analytics begins.
 Acknowledgments

T his book may not have been written but for sports and
politics—especially cricket with its focus on statistics and
unending debates on who is really great across different eras.
Both sports and HR are about people, talent, and contribution.
Nevertheless, one has the database for meaningful debates, while
the other is still evolving. This book owes to all the statistical
research studies on Gavaskar versus Tendulkar and so on.
We would also like to acknowledge the opportunities presented
by two of the organizations we had been associated in individual
capacities—Infosys and Indian Institute of Management (IIM).
The People Capability Maturity Model (P-CMM) framework and
the analytical rigor it called for in parallel with the HR scorecard
created enough opportunities to develop unique analyses. IIM,
Kashipur, offered an opportunity to connect the topic to the HR
community. Ramesh had worked with Sasken Communication
Technologies Ltd, where people were very receptive to use analytics
to review HR strategy.
We would like to thank our editor Sachin Sharma for staying
the long course, supporting the evolution across nearly three years
from a blind message on the website to a published book. While
one Sachin (Tendulkar) contributed to the causes, the other Sachin
(Sharma) enabled fleshing it out! We also thank Priya Arora and
her editorial team for diligently reviewing the book and converting
it into a final product.
The following people helped with their case studies, without
which this book would have been half done.

• Mr Richard Lobo and Mr Vinu Sekhar from Infosys

• Mr Saurabh Jain and Mr Neeraj Sanan from Spire2grow
xxii Winning on HR Analytics

• Mr Tej Mehta and his team from iCube Consulting Services

• Ms Tracey Smith
• Mr Mark Berry
• Ms Stela Lupushor
• Mr Srinath Thirumalai
• Mr Andrew Marritt
• Mr Steven Huang
• Ms Alexis Croswell
• Mr Ranjan Dutta
• Mr Eric Olesen
• Ms Gia D. Graham

Before we close, our thanks and acknowledgments to our fami-

lies. Hope their tolerance and unstinting support have been worth
the while!
 1
It Is the Right Time for Analytics in HR

CEO: Let us invest more in our people.

CFO: That is a risk! Their marketability will increase. What if
they quit?
CEO: What if we don’t invest in them and they don’t quit? Is it
not a bigger risk?

O ften when HR asks for more strategic involvement, it is asked

to show the evidence linking investments in human resources
of the organization to either top-line or bottom-line performance
or gaining competitive advantage. And this is where HR struggles
to find an answer. HR and corporate strategists are like prover-
bial rail tracks which have been struggling to find a meeting point.
While strategists are concerned about competition in the industry
and competitive challenges such as innovation, productivity, scal-
ability, customer centricity, etc., HR is more focused on ensuring
right talent at right time and right cost. Organizations repeat year
after year that people are their “key assets.” However, articulating
this asset value and appreciation tangibly has been tough even for
those organizations with best HR setups.

HR’s Tryst with Competitive Advantage

• In 1950s, Peter Drucker wrote, “Some wit once said mali-
ciously that [personnel management comprises] all those
2 Winning on HR Analytics

things that do not deal with the work of people and that are
not management.” (Drucker, 1954). And since then HR
has been struggling to be accepted as part of management
(or seat at the table). An article published by J. Barney, in
Journal of Management (1991), for the first time articulated
clearly on the resources an organization has and their link to
competitive advantage. The article built on resource-based
view (RBV) theory by E.T. Penrose (1959). RBV theory
has been seen as key in bridging the link between human
resource management (HRM) and business strategy. As per
RBV theory, any organization has tangible and intangible
resources. Tangible resources are land, machinery, or money
and intangible are goodwill, patents, or human capital pool.
Barney elaborated that resources can be sources of competi-
tive advantage only if they satisfy four criteria, namely the
VRIO framework:
• Valuable,
• Rare,
• Inimitable, and
• Organized.

Any resource—tangible or intangible—satisfying all the four

criteria can be a source of competitive advantage. A simple analysis
reveals that human capital pool is one resource which cannot be
easily imitated or may be unique (rare) to the organization and,
hence, has a huge potential to be the source of sustained competitive
advantage.

Human Capital Alone is Not Sufficient

Critics of RBV theory argue that having human capital alone is
not sufficient for competitive advantage. What is needed is a path
which facilitates interactions in the form of collaboration among the
human capital that leads to uniqueness and inimitability resulting
in competitive advantage. In layman’s language, behaviors, and
actions displayed by human capital at workplace are critical to
capitalize on this valuable resource. HR’s role becomes critical
 It Is the Right Time for Analytics in HR 3

in designing policies and procedures to encourage right behaviors

and actions delivering business performance.

HR Policies Are Critical Too

Human capital coupled with appropriate behaviors and supported
by HR policies creates a potent mix for sustained competitive
advantage. HR policies have not only to be aligned with the
organization life cycle stage and business challenges such as
productivity, innovation, scalability, etc., but also have an inter
se alignment. While the first type of alignment is called vertical
alignment or fit, the second type is called horizontal alignment or
fit, which was popularized by Lloyd Baird and Ilan Meshoulam
(1988). Presence of both the fits also ensures that the HR function
becomes complementary to other business functions in achieving
organizational performance. Vertical fit ensures cross-functional
collaboration between HR and other functions leading to better
appreciation of how HR contributes at the business strategy
level in solving key business challenges. Horizontal fit ensures
collaboration between various HR subfunctions so that their
synergy helps HR contribute in achieving business objectives.

Evolution of HR Approaches to Measurement Challenge

All the preceding arguments make one understand that human

capital plays a critical role in achieving business results. The
challenge then is to demonstrate a link between the HR, business
strategy, and performance using data. HR function’s tryst with data
is very old. Ever since the organized way of doing business started,
managers have been concerned with this cliché question—“How
to find the right person for the right job at the right time and cost.”
And answer to this is still evading managers.
Back in the early 20th century, a Philadelphia-based manufac-
turing company used a novel method to find the right people for
its various positions. This company would ask the potential job
4 Winning on HR Analytics

seekers to assemble as a group outside the company premises and

then the manager would toss the apple in the air. Whosoever caught
the apple amongst the group was offered the job!
Later on, after World War II, due to the acute shortage of
skilled employees, US Army started using skill tests to find the
people having right attributes and this was adopted by AT&T in
the corporate world. Subsequently more tests such as 16PF, TAT,
MBTI, and host of others were designed by various psychologists
to find the right people. With time, companies moved towards
competency-based practices that are based on attributes related to
workplace. Personality tests provided an optional supplement
to talent management processes.
Mid-1990s onwards company executives and HR function
started generating their own questions to find the right people, and
these included: Why manholes are round? How many triangles can
fit in a square of a particular size? etc.; these were more popularized
by product companies like Microsoft and Google which used these
questions while selecting people in 9–10 rounds of candidate
interviews.
HR measurement attempts so far have been confined to find
the right people for the right job and people who can deliver high
performance. However, in 1978, Jac Fitz-enz published an article
in Personnel Journal (the predecessor to Workforce Management)
titled “The Measurement Imperative.” In it, he proposed a radical,
anti-establishment idea: that human resources activities and
their impact on the bottom line could be measured. The response
received by Fitz-enz from HR practitioners was at best lukewarm
and cynical. However, this article had triggered debate and interest
by some other scholars leading to more publications on measuring
HR. This led to the beginning of data capturing for key HR
activities such as employee turnover, recruitment, compensation,
and training by the HR function followed by comparing data with
similar organizations in the same industry giving birth to what is
called “benchmarking.”
Thus began an era of benchmarking key HR measures against
the best practices which reached its peak in the 1990s and early
2000. But soon it was found that benchmarking was not providing
 It Is the Right Time for Analytics in HR 5

any insights for action and the only benefit was a solace how the
company was doing compared to others. Also during the 1990s,
there was emergence of human resources accounting and utility
analysis approaches to quantify human resources, but that had
limited impact.
However, in 2002 Oakland A’s use of metrics by its general
manager, Billy Beane, in the selection of team members and
subsequent publication Moneyball—The Art of Winning an Unfair
Game by Michael Lewis (2003) emerged as a path-breaking strategy
in the selection space. Oakland A’s with a paltry budget of USD
41 Million were competitive with teams with much larger budgets
like the New York Yankees. How A’s did this is very simple—
it extensively used sabermetrics (player data based on extensive
analysis of baseball) in the selection of players. The A’s found that
players with strong sabermetrics correlated to winning games than
those players who were strong in traditional metrics like batting
average used in the selection of baseball players. Also A’s found
that sabermetrics also offered an opportunity to put together the
match-winning team which was far less expensive. And traditional
metrics were used heavily by others while selecting their teams.
A’s challenged the established convention in selecting baseball
players and discovered that by using sabermetrics to measure the
player value, it got cheaper talent which delivered the results!
Extension of the Moneyball concept to the corporate world
happened in 2006; Billy Beane gave a talk on ‘Moneyball Approach
to Talent Management’ at an HR Conference in Texas, Austin, and
it caught the eye of corporate America. In 2009, Google started
“Project Oxygen” to find out “what makes a good manager.” In
year 2010, Davenport, Harris, and Shapiro (2010) published an
article in Harvard Business Review titled “Competing on talent
analytics,” thereby creating buzz in the corporate world. In 2011,
Google shared the results of Project Oxygen highlighting data-
based findings on what makes a “good manager”—forcing the
corporate world to take note of Google’s data-driven approach
to find attributes of a good manager. Soon thereafter there were
series of publications focusing on benefits of using analytics in
workforce or people management in Wall Street Journal, Forbes,
Harvard Business Review, Fortune, etc., including findings from
6 Winning on HR Analytics

Project Oxygen of Google (Garvin, Wagonfel, & Kind, 2013) that

showed that academic grades and types of questions it asks during
selection have no correlation to employee performance.
Corporate world saw a new hope and the possible answer
to the question of finding the “right people” by using a data-
based approach to workforce management which got labeled as
“workforce analytics” or “people analytics” or even “workforce
science,” while metrics-based HR analytics had been in use for
a long time in the corporate world. What workforce analytics
promised was a step beyond metric-based analysis to “predictive
analysis” using an algorithm-based model relying on huge volumes
of data (internal or external or often called big data) to make the
data-based people management decisions. In an interview to The
Atlantic, John Hausknecht, a professor at Cornell University
School of Industrial and Labor Relations, said,

In recent years the economy has witnessed a huge surge in demand for
workforce-analytics roles. You can now find dedicated analytics teams in
the human-resources departments of not only huge corporations such as
Google, HP, Intel, General Motors, and Procter & Gamble, to name just a
few, but also companies like McKee Foods, the Tennessee-based maker of
Little Debbie snack cakes. (Hausknecht, 2013)

What Is HR or People Analytics?

HR or people management has been traditionally seen as an
“art,” relying on the use of gut or intuition while making people-
or HR-related decisions in organizations. However, recent
developments, as mentioned earlier, have highlighted the benefits
of using data while making people decisions and thereby giving
a semblance of data-based objectivity (scientific basis) in people
decisions. This scientific approach to HRM in organizations has
given birth to a new field called HR analytics or people analytics
or workforce science, which uses a mix of understanding patterns
based on data algorithms and intuition in making people decisions
across an employee life cycle; typically, 80% data-based analysis
and 20% intuition seem to be the rule of thumb. It is generally
 It Is the Right Time for Analytics in HR 7

Figure 1.1 HR Analytics Continuum

Source: Authors.

defined as a systematic collection, analysis, and interpretation of

data to improve talent management decisions.
It is equally important to know what is not HR analytics or
workforce science. One view is that generally it does not include
metrics or dashboards, or reports of simple headcount or employee
engagement score or attrition data. Other view is that HR analytics
includes only predictive and prescriptive analytics. However, truth
lies in between. HR analytics is like a “continuum,” and on one
end it can be basic ratios and metrics and on the other it will be
complex algorithm-based predictive and prescriptive analytics. So
an organization can be anywhere on the spectrum based on the
maturity of HR processes, data quality, and capabilities available
(Figure 1.1).
With the dawn of data era, information is available in abundance
and low cost. Technological advances have facilitated the capture
of information across employee life cycle events, thus making
available humungous volume of employee data.

Why This Sudden Interest in HR Analytics?

For a long period, HR has been striving to get a seat on the table
along with finance, operations, and sales and marketing functions
to become a strategic function in any organization. In its quest to
become “strategic,” at the basic level, it has been demonstrating
its value-add to business by showcasing metrics focusing on
“efficiency,” such as lowering HR cost per employee or reducing
cost of per hire, etc.
8 Winning on HR Analytics

Some other organizations have gone a step ahead and showcased

“effectiveness metrics” such as employee engagement, satisfaction
increase, or employee retention increase to highlight HR’s value-
add to business.
However, the C-level has been skeptical of these metrics and
these have been generally labeled as metrics for justifying the
existence of HR without any tangible link to either top-line or
bottom-line performance. So this gap of showing how HR metrics
link to business metrics has always remained. HR needs to move
up the “measurement or metrics value chain” (Figure 1.2) from
efficiency–effectiveness metrics to “business impact” metrics to
demonstrate the link between HR metrics and business metrics.
These impact level metrics require the use of advanced statistical
modeling techniques and complex algorithms to perform key types
of analysis including predictive analytics, prescriptive analytics,
and cognitive analytics.
Predictive analytics will inform the C-level about “what will
happen,” for example, who will quit next, while prescriptive
analytics will inform the C-level about “what can be done to prevent
that attrition.” New generation of analytics like cognitive analytics
can identify patterns form large and complex data using multiple
hypotheses to identify patterns and insights which could not have
been seen earlier due to human limitations to construct models. For
example, cognitive analytics can convert a simple hypothesis into
a relationship between hiring channel and employee performance
into multiple patterns worth considering, which otherwise would
have required creating a hypothesis and analysis of data each time,
making the process akin to finding needle in the haystack. So this
kind of HR analytics, purely based on data, catches the attention of
the C-Level and, hence, provides an opportunity for HR to become
truly strategic, and this, in turn, will transform how HR is practiced.
Google leads in the use of predictive and prescriptive analytics
and lot of other large companies such as Shell, Procter & Gamble,
Morgan Stanley, Xerox, and General Motors have started using
these analytics. However, the number of companies globally using
these advanced HR analytics is very small. Latest study done by
Bersin by Deloitte in September 2013 shows that only 10% of
Fortune 500 companies are using these advanced analytics and
 It Is the Right Time for Analytics in HR 9

Figure 1.2 HR Analytics Value Chain

:KDWDFWLRQVFDQEH
WDNHQEDVHGRQSDWWHUQV
IRUIXWXUH"

Source: Authors.

out of this 10%, only 4% are using predictive and prescriptive

analytics, while other 10% are using basic statistical techniques for
HR analytics (Bersin, Leonard, & Wang-Audia, 2013). According
to Bersin (Fortune Magazine, March 21, 2016), the number of
companies using predictive analytics has risen to 8%. The major
reason why only a small number of Fortune 500 companies are
using HR analytics is because HR faces big challenges to scale up
for using HR analytics.
10 Winning on HR Analytics

Big Data Era and HR Analytics

Everything which is connected today to Internet is generating
volumes of data in one form or another from various sources such
as handheld devices, laptops, and machines. This generation of
voluminous data was characterized by Laney (2001) as “3D data”
meaning data volume, data velocity, and data variety. Later on,
some organizations added “data veracity” to it, making it 4 Vs.
More Vs such as value, variability, and visualization have been
added by some organizations to enhance the character of big data.
Big data is positioned as equivalent of telescope and microscope
in terms of revolutionary ideas. As telescope helped to see stars
and galaxies never visible to naked eye in the universe, similarly
microscope made it possible to capture the life at cell level. In
the same manner, big data is helping businesses to capture unseen
patterns and trends, find answers to unanswered questions, and
deal with uncertainty.
Today big data is being used in sports, politics, retail, insurance,
healthcare, public departments like police, etc., for making deci-
sions which earlier were made on the basis of intuition. Within an
organization, functions such as marketing, finance, procurement,
etc., are using volumes of data to streamline business processes
and increase organization’s efficiency and effectiveness. However,
there is a limit up to which factors of production, like machine, can
be maximized for efficiency while there are several blocks related
to another key factor of production—human factor—which prevent
any efficiency and effectiveness optimization. And “human factor”
is the key driver of business in service and other people-intensive
organizations. There have always been attempts to align human
factor of production with organization performance through vari-
ous methods such as better training, rewards and recognition, work
redesign, etc., but with limited success. As it is said that organiza-
tions do not produce business, it is the people or humans working
in an organization that do so.
Traditionally, HR has always been collecting volumes of data
on various dimensions of human resources, such as:
 It Is the Right Time for Analytics in HR 11

• Demographic
• Performance management
• Compensation/benefits
• Educational history
• Job location
• Training
• Talent movement

However, this data has been used so far to compute metrics

or ratios and do benchmarking. As highlighted earlier, bench-
marking helped an organization to compare its status of HR
practices vis-à-vis others in the industry and get an idea about
various best practices followed by the industry. For example,
if an organization had an attrition rate of 17% and the industry
attrition rate was 19%, the comparison only gave solace to the
organization that it is doing better than the industry but did not
inform it about who is leaving, why they are leaving, and what
can be done to prevent attrition. Thus, organizations were merely
using available data to justify its existence by comparing with
industry, whereas board and CEO wanted HR to show evidence
of HR investments impacting the top-line and bottom-line.
Cascio and Boudreau (2011) pointed out that HR faced “big
wall” in moving beyond benchmarks and scorecards to demon-
strate evidence-based HR’s impact on revenue and profitability
(Figure 1.3). In Figure 1.3, the vertical column is the “wall” some-
times called china wall for HR, which HR has been attempting to
cross to demonstrate its “strategic impact” in terms of causation
and organization effectiveness.
Big data’s HR promise is to help cross this wall and demonstrate
its impact on top-line and bottom-line by showing with data how
various elements of employee life cycle can drive revenue and
profits and make “people” as a source of competitive advantage
and become truly strategic like other business functions. In the
next few chapters, we will explore how data-driven HR can help
demonstrate its business value and become truly strategic.
12 Winning on HR Analytics

Figure 1.3 Scope HR Measurement Approaches

Source: Cascio and Boudreau, 2011.

Business Strategy–HR Analytics–Competitive

Advantage Integration
A 2011 MIT study on analytics found that top performing
organizations use analytics five times more than the lower
performing organizations (LaValle, 2011). HR analytics can help
organizations to deal with competitive landscape at tactical and
strategic levels. At strategic level, typical competitive challenges
faced by any organization include productivity, innovation, global
scaling, lean delivery, etc. The HR function can align vertically and
horizontally using a data-based approach to help an organization
deal with these competitive challenges and gain competitive
 It Is the Right Time for Analytics in HR 13

advantage. Figure 1.4 provides an overview of business challenges

and HR analytics linkage. In Chapters 2–9 of this book, we will
cover how HR analytics and right metrics can help HR to provide
evidence of contributing to organization performance at the
strategic level.

Figure 1.4 Competitive Challenges–Business Strategy–HR Analytics

Integration

Source: Authors.
 2
Articulating Business Value of
HR Programs

E conomy has evolved from purely agrarian based to industrial to

information and now to idea or intelligence based. In keeping
with this, HR has shifted from a pure policing–administration
role to personnel to business partner and now to key player at the
strategic level. This shift has been largely due to recognition of
people as resource with a pair of hands to people with ideas as key
players in execution of business strategy and source of competitive
advantage. Publications like McKinsey’s classic War for Talent and
series of research studies in the last two decades have contributed
in highlighting the role of HR in business strategy.
In many organizations, cost of payroll and running various HR
programs amount to as high as 50% to 75% of organization’s total
costs. Hence, it is no surprise that a reduction of 10% in HR costs
results in significant uptick in net profits, while cutting cost is the
easiest way to shore up profits with the other option being for HR
to showcase how investments made in various HR programs create
business value and can even create greater value by increased
investments.
In a predominantly knowledge- and idea-based economy,
studies have shown that intangibles have 30% to 40% impact on
market capitalization of the company.
 Articulating Business Value of HR Programs 15

• A study by Ernst & Young Center for Business Innovation

(Mavrinac & Siesfeld, 1998) found that intangible factors
(e.g., strategy execution, managerial credibility, strategy
quality, attracting and retaining talent, management experi-
ence, and compensation strategy) explain much of the vari-
ance in the market value of companies. The impact of these
factors varies across industry; for example, in the technol-
ogy industry, the quality of management explains as much
as 13% of the total variance in market capitalization.
• A large study by Huselid (1995) found that companies with
sophisticated HR programs/systems (also known as “high-
performance work systems”) have a significant financial
impact on profits per employee, sales per employee, and
market value per employee. Findings of this and other similar
studies have gained the attention of executives interested
in measurement of HR programs as well as in redesigning
employee programs like appraisals to ensure that HR is held
accountable for enhancing their workforce’s contribution to
the bottom-line.

A Conference Board survey (Gates, 2002) found that HR’s efforts

to showcase its business contribution have been facing challenges
on the measurement front as the C-level was not convinced of the
measurement effectiveness as compared to marketing, finance,
or operations. However, with the advancement of technology,
falling cost of data storage has come as boon to HR to tide over the
measurement challenge. In the last decade, IT applications have
linked dispersed data in HR subfunctions to get data insights into
various HR subfunctions. This emerging field of HR analytics has
bridged the gap of reliability in HR measurements. Today, it is no
longer question of having no faith in HR measurement effectiveness
but of where, what, and how much should be measured which
shows HR’s impact on business outcomes.
16 Winning on HR Analytics

HR Analytics Linkage to Business Outcomes

HR analytics can impact business outcomes such as sales,
productivity, profitability, customer satisfaction, etc., either by
adopting an HR measurement system covering all HR practices or
by focusing on a single HR practice, like recruitment, without a full
measurement system.

1. Adoption of HR analytics as a model or framework

using technology/tool: Here, the focus is on how the
implementation of technology- or system-based HR analytics
as a cogent model or framework of some maturity level by
any organization impacts business outcomes, for example,
use of a human capital management (HCM) reporting and
analysis system.
2. Adoption of a HR program/practice: Here, the focus is on
how the use of various HR programs such as employee
engagement surveys, workforce planning, succession plan-
ning, compensation management, performance manage-
ment, learning and development, etc., based on analysis link
to business outcomes. For example, based on recruitment
data analysis, an organization identifies certain attributes
linked to high performers and then hires those who have
those attributes. Some of the programs will be covered in
this chapter while others are covered in subsequent chapters.

Measuring Use of HR Analytics Impact on

Business Outcomes
HR function has been collecting various types of employee data
for many decades now. Much of this data by any organization
has been used to gather inputs on what has happened in the last
year and what is happening at the moment on HR metrics, such
as headcount, attrition rates, absence rate, cost per hire, training
hours per employee, etc., primarily to get a sense of the current
HR temperature in the organization. Increasing adoption of HCM
 Articulating Business Value of HR Programs 17

suites by organizations enabled them to play with core HR data

and HCM suite’s data and, thus, marked the beginning of use of
HR analytics at the organization level. Now an organization is able
to track which units or projects are having more usage of training
resources or where performance appraisal quality is better than
other units or projects. As acquisition of HCM suites came with a
substantial cost, management became interested to know how the
use is impacting business outcomes.
Aberdeen Group published a study in 2012, based on an extensive
study, to find out the impact of deployment of HR analytics on
business outcomes by comparing data of organizations which
had adopted HR analytics to support business strategy versus
those which were not using it (Lombardi & Laurano, 2012).
The study found that the use of HR analytics helped companies
to achieve higher results in the range of 8%–15% for customer
satisfaction, customer retention, and revenue per employee as
shown in Figure 2.1.
Another study done by CedarCrestone in 2014 again found that
those companies which had adopted HR analytics technologies
outperformed on profit and revenue per employee parameters
all those which had not adopted the same (Figure 2.2). The

Figure 2.1 Business Impact of HR Analytics

Source: Aberdeen Group (Lombardi & Laurano, 2012).

18 Winning on HR Analytics

Figure 2.2 Financial Performance Growth by Talent Management

Technology Adoption

Source: CedarCrestone HR Systems Survey (CedarCrestone, 2014).

difference is 16% on revenue per employee and 35% on profit per

employee.
Hence, consistent results by the above-mentioned two
studies indicate that the use of HCM reporting in some manner
or performing analytics in some form results in better business
outcomes than those organizations which do not use it.

Measuring HR Programs for Business

Results Linkages
Companies often wonder to find an answer to the question “what
makes a successful company? Is it buildings or perks or infrastruc-
ture or culture or people?” And this quest for answer has resulted
 Articulating Business Value of HR Programs 19

in many studies pointing out that everything being equal, people

are the key differentiators. Almost all the organizations deploy
variety of HR programs across the employee life cycle starting
from attracting talent to exit of talent and even in post-exit situa-
tions through alumni programs. All of these programs cost millions
of dollars to companies. For example, companies are spending
close to USD 1.5 billion in employee engagement program/survey
(Kowske, 2012) and yet employee engagement across the globe is
mere 13% based on a 142-country survey of employee engagement
levels (Gallup, 2012). Hence, it makes sense for managements to
demand contribution of investments in HR programs like those for
employee engagement, for better business outcomes.
Let us explore with the help of a hypothetical case. If an IT
company decides to spend `10,000 per employee per annum on
various HR programs, the total spend for different sized companies
will be as shown in Table 2.1.
Let us look at HR programs investments from management’s
lens with a focus on net profits/earnings before interest, taxes,
depreciation, and amortization (EBITDA) (Table 2.2).
Now let us assume that the CEO and CFO of Company A tell the
HR head that they are doing away with the HR spend on training
worth `10 crores for this year to increase earnings per share
(EPS) so that it increases stock price to attract investors for more
investment for future expansion plans. Alternatively the HR Head

Table 2.1 Sample Spends for HR Programs

HR program
Revenue Total HR spend as
(approx.) program percentage of
Company Headcount (`) spend revenue

A 10,000 2,500 crores 10 crores 0.40

B 20,000 5,000 crores 20 crores 0.40

C 50,000 12,500 50 crores 0.40

crores

Source: Authors.
20 Winning on HR Analytics

Table 2.2 HR Programs Spend and EBITDA Linkage Analysis

HR
HR program
Net Profit/ program spend as
Revenue EBITDA Total HR spend as percentage
(approx.) @ 20% of program percentage of
Company Headcount (`) revenue spend of revenue EBITDA

A 10,000 2,500 500 10 crores 0.40 2.00

crores crores

B 20,000 5,000 1,000 20 crores 0.40 2.00

crores crores

C 50,000 12,500 2,500 50 crores 0.40 2.00

crores crores

Source: Authors.

can continue the training spend if there is an evidence that such a

training investment has an impact on business outcomes (increased
sales or profits).
Company A is listed on Bombay Stock Exchange (BSE)
and National Stock Exchange (NSE) and has an outstanding
share capital of `200 crores with `10 face value share and the
total outstanding shares as 20 crores. Assuming that the current
market price of share is `250 and with 500 crores EBITDA,
the EPS will be `25 giving a price to earnings (P/E) multiple of
10 (250/25).
Assuming if CEO and CFO do away with the HR training spend
of 10 crores per annum, new EBITDA will be 510 crores and new
EPS will be `25.50. Assuming that P/E remains 10, new stock
price will be 255, representing an increase of 2%.
Continuing with Company A, the typical total employee cost will
be 60% revenue or 1,500 crores. Assuming that the management
of Company A decides to reduce the overall HR budget/employee
cost by 5%, then this reduction of 5% in 1,500 crores can make
the EBITDA 575 crores, and assuming that the training spend of
10 crores continues, new EPS with 5% overall reduction will be
28.75, and assuming P/E of 10, new stock price will be 287.5, an
increase of 15% over 250.
 Articulating Business Value of HR Programs 21

The above-mentioned simple case study explains the point that

reducing cost or cutting HR cost by either reducing the number
of employees or reducing the training spend, etc., is the easiest
way for HR to show proof of impact on the bottom-line. However,
there are other cost levers too apart from HR budget or workforce/
employee cost reduction which can impact business outcomes like
profitability, and the HR function can play an important role in
driving those cost levers to impact business outcomes. And that
is what the HR function should focus on—using data to showcase
how it impacted business.
For example, cost of production can be reduced by capacity
expansion to impact business outcomes, and the HR function can
facilitate this. In a manufacturing setup, it means increasing the
current operating capacity closer to the installed capacity without
increase in labor/manpower, leading to increased output with the
same labor inputs. It can be easily measured to show impact on
business outcomes. In a service or IT company, capacity expansion
is possible by increasing the utilization rate of deployable or
billable employees, again thereby increasing revenue with the
same resources at the same cost. These are some of the easier ways
to link HR programs, like training investments resulting in capacity
expansion and, hence, higher revenue. And majority of companies
employ these methods to increase revenues. However, the HR
function needs to articulate its role in such methods of increasing
revenues rather than easily accepting reduction in HR budget or
headcount to reduce costs and increase profits.
Another measureable tactic, other than cost reduction, which
HR can use is to boost profitability through use of technology.
Many companies are using technology to provide self-service
to employees enabling them to pursue e-learning, get answers to
queries like compensation, manage benefits and performance,
etc. On an average, efficient deployment of technology for HR
administrative processes can reduce cost up to 30%, which is quite
substantial and measurable also.
There are other ways of showcasing impact or contribution of
HR on business outcomes by increasing revenue or sales. This can
be illustrated with the help of an example as below.
22 Winning on HR Analytics

Let us assume that company A grows by 20% and new revenue

is 3,000 crores, which increases wages by 5% and continues with
10 crore HR training investments; EBITDA at 20% will be 600
crores and EPS will be 30 and with P/E of 10, new stock price will
be 300, an increase of 20% of over 250.
So in the second case, if HR can show evidence that the increased
revenue is due to increased investments in training of people of the
company, then reduction in such investments becomes agnostic to
the market price of the share for a listed company.
Let us take another illustration to explain the difference between
cost levers and engaged and developed people as a lever for
impacting business outcomes. Take two companies A and B with
below data points as shown in Table 2.3.
Company A has a high engagement score and higher revenue
per employee, but higher cost per hire. Company B has lower
engagement score, low revenue per employee, and lower cost per
hire. Here, company B’s HR is very efficient by keeping hiring
costs low while company A has a higher hiring cost. Comparison
shows that HR is saving money and adding to profits in company
B while people/talent is driving higher revenue per employee in
Company A through high engaged and developed employees. In
both the situations value gets created, though through different
methods, but higher revenue per employee has more competitive
advantage than the cost reduction approach. If a company can do
both, then there will be a multiplier impact on outcomes.
Like employee engagement, HR has lot of levers to pull for
impacting revenue or profitability through investments in HR
programs which can help in building value over time and provide

Table 2.3 Sample Employee Engagement and Revenue Link

Employee Revenue per

Company engagement score Cost per hire employee

A 37 50,000 11,00,000

B 31 34,000 9,50,000

Source: Authors.
 Articulating Business Value of HR Programs 23

sustainable competitive advantage. These include building a

culture or work environment which helps in attracting the right
talent, developing and rewarding it, and facilitating the employees
to perform at their best. This is where HR needs to invest time and
money, and measurement can help in identifying where maximum
investment can provide maximum returns.
However, challenge still remains in measuring the impact of
a large number of HR programs on business outcomes. Some
research evidence is there which has attempted to find evidence of
HR program’s impact on business outcomes.

Research Evidence on Impact of HR Programs

Cantrell and her team conducted a study of several companies
in 2006 to find out the linkage between specific HR programs
in business organizations (Cantrell, Benton, Laudal, & Thomas,
2006). Study found that companies which had a focused approach
towards the following HR processes had achieved greater economic
success than those which had lesser focus on these processes
(Figure 2.3):

1. Process aligning people strategy with business strategy,

2. Process providing supportive work environment to
employees, and
3. Process for developing employees by giving them ample
opportunity to learn and grow.

Analysis showed that improvement of one quartile in these

processes resulted in 10%–15% increase in capital efficiency (ratio
of total annual sales to invested capital).
Another study done by Conference Board (Gates, 2008) on
use of HR analytics by Marriott Vacation Club found that sales
executives with higher engagement score in the top quartile had
delivered higher performance on parameters like sales volume
per guest and percentage of tours losing on time-share contracts
(Figure 2.4).
24 Winning on HR Analytics

Figure 2.3 Superior HR Processes and Financial Performance

Source: Cantrell et al. (2006).

Notes: a Financial results are represented by the two year average of organization’s
capital efficiency, or the ratio of total annual sales to the capital invested in the
operations of the business by shareholders and creditors.
b
 Human capital process benchmarking rankings are based on relative perfor-
mance of effectiveness scores compare to other scores in the database. Ranking
reflect quartiles of performance (4–5 = top quartile, 3–4 = second quartile, 2–3
= third quartile, and 1–2 = bottom quartile).

How to Measure Linkage of HR Programs to

Business Outcomes?
HR practitioners know the value of measuring how HR programs
are contributing to business outcomes, yet often struggle with the
“how” part. Let us use some illustrations to get the understanding
of the “how” part.
Engagement survey or job satisfaction survey is one of the major
employee initiative or program for any company. It is done largely
on annual basis and covers large number of job variables impacting
 Articulating Business Value of HR Programs 25

Figure 2.4 Highly Engaged Sales Executives Result in High Productivity

Source: Stephen Gates (2008).

employee performance and associated factors such as culture,

leadership quality, company brand, etc. Data collected through
such survey are large and rich and subjected to all possible ways
of data dissection and bisection. Analysis of data is done in a very
detailed manner but unfortunately such an analysis does not result
in any meaningful actions after the survey to address employee
engagement issues. And this lack of impactful action based on
survey has led to it becoming a ritual affair and also employees not
giving much credence to it. There are several reasons why analysis
ends up in “action paralysis,” but one of the major reasons is no
attempts made by HR leadership of the company to survey findings
with business challenges or to explain how survey provides insights
for action based on the data analysis.
26 Winning on HR Analytics

There are many ways in which the HR team can play with
engagement survey data and link it to business outcomes, for
example, in the case of an IT company which wants to know how
the survey data can help in finding out the impact of software
developer engagement and attrition on business outcomes.
For doing this linkage analysis, the following easy steps can be
followed:

1. Create a model showing linkages between dependent and

independent variables/factors (Figure 2.5). Authors have
created this model based on review of the literature on
employee engagement and employee attrition.
2. Identify sources and types of data required for performing
such an analysis.

Figure 2.5 Measuring Impact of Software Developer Engagement on

Business Outcomes
 Articulating Business Value of HR Programs 27

3. Perform statistical analysis based either on regression

analysis or structural equation method (SEM).
4. Identify the patterns based on data analysis.
5. Draw insights based on patterns.
6. Finally, identify actionable steps.

Another key HR program/initiative is learning and development

as this is a large component of the operational budget. Often
questions are raised on what is the return on investment (RoI) of
training programs and why not to do away with these. Measurement
of training programs for business impact has been a challenge
primarily due to the nature of data used in measurement. So far
training measurement has been done well at two levels—efficiency
and effectiveness. Typical measures used to capture efficiency
have been attendance rate, cost per participant, etc. Similarly,
effectiveness measures used include knowledge acquired test,
knowledge applied on the job, etc. However, both these types
of metrics, though relevant, fail to provide evidence of business
impact.
Let us take the case of a large automobile company in the
passenger car business with 1,000+ dealers across the country
and about 1,000+ sales managers (SMs). On the basis of dealer
sales analysis, it found that 30% of SMs working with dealers are
performing at an average level. Further internal studies showed
that these SMs not performing as desired were lacking in key skills
related to the SM role. Company decided to roll out a 2 day sales
training program for 300+ SMs across the country. At the same
time, the company wanted clear evidence of training program
impact on business outcomes from HR.
So how should HR go about showing business impact of
training in the above case? To begin with, HR will have to identify
what data is available within HR, and this will include employee
demographics, annual performance review data, employee
education, and training history data for SMs. Other data needed
by HR will be related to business outcomes or key performance
indicators (KPIs) for SM covering car sales volume, gross profit,
28 Winning on HR Analytics

inventory turn, productivity, and dealer satisfaction score (DSS).

For KPIs-related data, HR will have to work with business
managers, finance function, production function, marketing
function, etc., as they are the custodians of business data. The data
in possession of HR will not suffice for actionable analytics work.
The model for linking training to business outcomes will look as
follows:

Next step will be of identifying SMs who should go for training

based on SM data using criteria such as tenure, performance
ratings, training history, etc. Along with this data, business
outcome data for each SM before training participation needs to
be collected. After going through the training program, business
outcome data needs to be collected again after a period of 2–3
months to capture training impact on business outcomes. At the
same time, the company can compare this data with performance
data of SMs not covered by training program. Improved business
outcomes for those who completed their training over the non-
trained SMs clearly provides evidence of impact on business
outcomes. As all the business outcomes can be quantified, it is
easy to add monetary value to improvement in business outcomes
due to SM training.
Similarly, actionable HR analytics for linking other HR
programs such as attrition management program, retention
program, incentives programs, etc., can be done with the help of
data available within HR and other business functions. As HR
professionals are increasingly being asked to demonstrate HR
programs on business outcomes, the criticality of HR analytics
becomes more important.
 Articulating Business Value of HR Programs 29

Industry Examples of Measuring HR

Programs Impact
Sears Inc. measures impact of retail store employees’ engagement
on store performance. It captures data on “Moodrings” which
shows how the mood of an employee changes through the day
and impacts customer interaction and experience. It has crunched
1 million data points and found that an employee in good mood
results in better customer experience leading to higher customer
sales and repeat sales. As a next step, Sears is identifying the
factors related to employee “good mood” and is focusing on those
to enhance store performance.
Disneyland measures impact of “smile” on customer satisfaction
and repeat visits by customers.
ConAgra used HR analytics software to predict which key
employees are most likely to leave the company and why. The
company scrutinized operations with high turnover and those with
low attrition, and mined data looking at over 200 factors that might
contribute to employees leaving. Employees’ relationship with
their supervisor and whether or not they were recognized for their
work were two of the stronger predictors.
Xerox services HR analytics for hiring its call center employees
as it was facing 60% attrition rate. Data analysis of online
personality testing for selection showed that previous experience,
one of the few criteria that Xerox had explicitly screened for
in the past, turns out to have no bearing on either productivity
or retention, while distance between home and work is strongly
associated with employee engagement and retention. Using
these parameters, Xerox reduced its attrition from 50% to 40%
resulting in savings of millions of dollars.
Human Dynamics Laboratory at MIT in recent years has pio-
neered a concept called Sociometry with the use of specialized
electronic “badges” that transmit data about employees’ interac-
tions as they go about their days. The badges capture all sorts of
information about formal and informal conversations: their length;
the tone of voice and gestures of the people involved; how much
those people talk, listen, and interrupt; the degree to which they
30 Winning on HR Analytics

demonstrate empathy and extroversion; and more. Each badge

generates about 100 data points a minute. Badge usage was tried
out on about 2,500 people in 21 different organizations, and a
number of interesting findings were discovered. Study found that
about a third of team performance can be predicted merely by the
number of face-to-face exchanges among team members. Using
data gathered by the badges, it can be predicted which teams would
win a business-plan contest, and which workers would (rightly)
say they had a “productive” or “creative” day.
All of above examples indicate that every organization wants to
improve efficiency and effectiveness of factors of production with
special focus on human as a key factor of production. Organization’s
eternal quest for data-based efficiency and effectiveness for driving
business results resulted in the birth of Taylorism and Fordism
at the early part of the 20th century, pioneering the use of stop
watches on shop floor. This was followed by assessment methods
to find the right employee for the right job based on data rather
than intuition around the mid-20th century. With the advancement
of technology and simultaneous growth in data to zeta bite levels
making easier for organization to deep dive inside the organizations
and individuals to draw insights on what works best for employees
to give better business results and competitive advantage. And this
use of “data intelligence” in people management is likely to grow
in the coming years as organizations realize increased business
value by making people decisions based on data insights.
 3
Analytical Problem Solving

W e have seen different frameworks for analysis in Chapter 1.

At the same time, when the business needs to solve a
problem, we cannot say that “We are in level 2 maturity and that
is a level 4 problem. We will solve it when we get there.” We need
to follow some approaches for solving problems using analytics.

Deep and Wide Approach

Let there be two companies. Company A has an attrition of 16%
and company B has 18%. They are in the same business and have
a similar size. In this situation, who is doing better?
One would, on the basis of data, say that A is doing better.
However, let us examine the following graph.
It is apparent from the graph on the next page that taking the
rates from a point in time alone may not be adequate. One needs
to take the trend across a longer period of time. A is showing a
downward trend, while B is showing a rising trend. Timeframe
becomes a factor too.
We also need a frame of reference. Let us consider two compa-
nies with attrition rates as in the table on page 32.
It is evident that company 1 is having an attrition rate that is less
than half of company 2. Is company 2 doing something wrong?
32 Winning on HR Analytics

2013–14 2014–15

Company 1 16% 18%

Company 2 42% 37%

Now, let us add another comparator. What is the attrition rate in

the industry in which the company is operating?

2013–14 2014–15
2013–14 Industry 2014–15 Industry

Company 1 16% 13% 18% 12%

Company 2 42% 65% 37% 70%

This table changes our viewpoint completely. While in the

absolute, company 1 is much better than company 2, when we add
industry number, it is company 2 which seems to be doing better!
Company 1 is not only having attrition rates higher than the
industry, but the rate is also increasing when it is coming down for
the industry. On the other hand, company 2 has lower than industry
rates and it is showing a declining trend, while the industry is
moving up.
 Analytical Problem Solving 33

One might be surprised with such high rates of attrition.

However, very high attrition rates are common in industries where
the hours are irregular and working conditions can be difficult
(such as call centers and retail). A product development company
might collapse if it loses people at 40% as the institutional
knowledge would be lost irrevocably. So their retention needs to be
far stronger. On the other hand, the processes are standardized and
anyone can do the job with some training when it comes to stacking
groceries in a shop. The cost of attrition is more manageable.
Industry benchmarks are sufficient to say how well one is doing.
What about setting the direction? What if the company is already
doing well against the industry?
Companies aspire to become employers of choice. To achieve
this, they need a benchmark. Pan-industry surveys come in handy
in this. There are different studies, such as “Great Places to Work,”
and “Best Employers”. Participants in each of this not only are able
to compare themselves to a norm but also can see if they are best
in class or what improvements are needed to become best in class.

Engagement
Company Industry Company Average score for
attrition attrition engagement engagement employers of
rate rate score score choice

13% 15% 67% 67% 78%

The data shows that the company is doing a good job of retaining
more people when compared to competition. The engagement
levels of the company are comparable to the participants in the
employer of choice study. However, there is an 11% gap with the
companies rated as “Employers of Choice.” So, there is more to be
done by the company in the engagement space and that would, in
turn, improve its retention.
Summarizing, a company can compare its performance using
the following:

• Historical trends
• Internal benchmarking across business units
34 Winning on HR Analytics

• Industry benchmarking
• Pan-industry benchmarking data using workplace surveys
• Global benchmarking using surveys, benchmarking ex-
changes, and best practices sharing.

Now, let us look at a different challenge. Your company has an

offer conversion rate of 70%. The benchmark for the industry
is 85%. You would like to improve it to industry levels. Your
recruitment team feels that having a sign-on bonus would help
improve the conversion rates. How would you go about deciding
on a course of action?
First of all, it would help to identify the problem. Is the join
rate 70% across the board? Unlikely. First step would be to split
into two categories—entry level and experienced. You might get
a figure as follows.
Entry level conversion rate: 90%
Lateral hire conversion rate: 64%
It is obvious that the join rate is impacted by having only 2 out
of 3 people who were offered actually taking up the offer. It makes
sense to focus on increasing the conversion rate of laterals.
How do you analyze the lateral join rate? You could do it on the
basis of the following:

• Functions,
• Job levels/experience levels,
• Roles,
• Locations.

Let us go one by one.

Function R&D Operations HR Infrastructure

% lateral hire 55 25 3 2

Offer conversion 60 75 50 90

Note: The highlighted columns indicate the segment where the conversion is below 64%.
 Analytical Problem Solving 35

HR has the lowest offer conversion rate. It is something that one

needs to take an action on. However, only 3% of all offers are made
for HR. Even a join rate of 100% in HR may not make a dent on the
company’s number.
The focus areas then are operations and R&D. Especially the
latter, as it has the highest proportion of new joins and a lower rate
of conversion. To begin with, let us take R&D. Is the joining rate
different for different locations?

Location Bangalore Pune Gurgaon

% offers 40 30 30

Join rate 55 65 60

There are some variations based on the location. At the same

time, the differences are not marked across locations. Just focusing
on Bangalore may not bump up the conversion rate. We then look
at the next dimension—experience levels.

Experience level years 2–4 5–7 8–10 11+

% offers 55 30 10 5

Join rate 60 55 90 90

Note: The highlighted columns indicate the lower end of the comparable demographic.

There is nothing unusual in this. Conversion rates are typically

lower for the levels where competition for talent is highest. Let us
drill it down a further level and see if there is a need for a sign-on
bonus, at least for the 2–4 year experience bracket.

Compensation quartile Q1 Q2 Q3 Q4

% offers made 35 35 20 10

Offer conversion rate 35 50 75 90

Note: The highlighted columns indicate the lower end of the comparable demographic.
36 Winning on HR Analytics

Finally, we have been able to drill down the problem. One-third

of the offers being made for people in the 2–4 year experience
bracket are in the first quartile of the compensation band. The
conversion rate here is only 35%. This pulls down the conversion
for people in this experience category who constitute 55% of all
offers made for R&D, which in turn accounts for 55% of all lateral
hiring.
The company has been making offers at the lower end of the pay
range. While this is a cost-effective solution, it has resulted in offers
being made but not having commensurate impact on the joining. A
sign-on, like what is recommended by the recruitment team, may
just be a band aid. HR needs to relook and come out with a better
way of assessing and fitting new hires equitably. The distribution
is skewed towards the lower half; 70% of offers are made below
the median and only 30% above. Given that conversion rates are
higher in the upper half, the company can analyze the reason for
low salary and explore ways to make an equal distribution across
the board. That is simplistic, but suffices for the purposes of this
illustration!
In the first situation, we explored the width of benchmarking,
standards, etc. In the second, we drilled deep into the numbers to
arrive at the solution. Analytical problem solving involves both
width and depth. Effective problem solving needs a judicious use
of both.
 Analytical Problem Solving 37

Building the Cube

We have used the deep and wide framework for visualizing
analytics. This is composed of multiple scenarios of two-
dimensional analysis. As the example shows, we have taken the
attrition levels of two companies and compared them; this analysis
is useful and a great starting point.
Now, let us consider two divisions of a same company and track
their headcount growth.

Division Headcount 2014 Headcount 2015

Division 1 1,500 1,700

Division 2 1,800 2,500

The inference is clear. Division 1 has grown its headcount by

13%, which is a decent number. On the other hand, division 2 has
registered an impressive growth of 38%.

Let us pose the next question. Has there been a difference in

terms of how the two divisions have grown their headcounts? In
which case, we need to look at their structures. Let us have a simple
structure of four levels, namely engineer, designer, manager, and
general manager. Let the composition before and after be as follows:
38 Winning on HR Analytics

Division 1 Division 1 Division 2 Division 2

Level 2014 2015 2014 2015

Engineer 1,200 1,275 1,435 2,000

Designer 200 280 240 330

Manager 80 115 100 135

General Manager 20 30 25 35

Our first level of drill-down showed differential rates of

growth. However, one would also need to understand the quality
of headcount growth. As the enclosed graph shows, the staffing
pyramid has remained constant for division 2, while for division
1 the number of GMs and managers has shown disproportionate
growth. It is possible that while division 2 is business growth as
usual, division 1 is actually looking to expand business, hiring
more senior people to seed/grow new businesses.
 Analytical Problem Solving 39

A graph based on relative percentage lays it out a little more

clearly. The proportion of engineers has come down from nearly
80% to 75% for division 1. Immediately, we are able to grasp
that something fundamentally different is happening there, which
would not have been visible with the two-dimensional graphs.
This three-dimensional analysis of building the cube helps in the
following illustrative scenarios.

• Assessing effectiveness of recruitment: Let us look at the

interview offer conversion rates. Typically, it is felt that the
higher the conversion rate, the more productive the talent
acquisition. A company may be having a conversion rate
of 75% and another having 85%. Next question would be
the trend. Let us say both have improved by 5% in the last
12 months. Both have made 1,000 offers each. So who is
doing better?
Let us examine the data further, by breaking it into offers
made for “freshers” and offers made for “laterals”. We find
that the company having a conversion rate of 75% actually
has a 60/40 ratio of selection. 40% of their offers are for
laterals. The second company has a ratio of 85%/15%. In
general, the conversion rates for freshers are much higher
40 Winning on HR Analytics

compared to lateral joins. So, even though the first company

makes 25% more lateral offers (40% compared to 15%) their
overall conversion is reduced only by 15%. For an equal
number of offers and other things being equal, the first
company is doing a better job of acquiring talent. The three
dimensions we have used are:
• Conversion rate,
• Breakdown into lateral and fresher conversion rate,
• Performance across 2 years on the conversion rate
measure.

Visualizing the required information in different dimensions

helps us get to the solution faster, when using analytics.

Going from Opinion to Insight

The deep and wide framework acts as a reference for any analysis
that we do. However, the human mind does not follow logic
always. As the example discussed above shows, often we arrive
at possible options for decision making based on intuition. But
decisions cannot be always taken on the basis of experience and
intuition alone. One needs to validate the opinion using a structured
approach and arrive at the right insight.
Let us take a simple example. HR teams are always looking for
ways to improve employee engagement. This in turn drives what
 Analytical Problem Solving 41

are called “fun events.” At the same time, one does not really know
at what frequency the fun events improve engagement levels and at
what level do they stop having an impact.

Opinion
We start with an opinion that says “fun events improve employee
engagement.” To validate this, we need to collect data.

Data

What is the kind of data one would need to collect? It has to be

around fun events on the one hand and engagement numbers, on
the other. We widen the scope of data collection.

• Employee engagement scores at two different points in time,

• Number of fun events preceding each engagement survey/
dipstick,
• Count of all fun events,
• Division of fun events into major/minor, across different
locations, across different business units,
• Number of people participating in different fun events,
• Retention rates of people participating in engagement
programs,
• Attrition on a monthly basis.

Once we have collected all this data, we organize it based on

metrics.

Metrics

The broad areas of metrics for analyzing would be as follows:

• Engagement levels; absolute and movement

• Frequency and number of events; absolute and movement
• Retention of employees.

With these, we move onto analysis.

42 Winning on HR Analytics

Analysis

We now need to analyze to get insights. Some threads can be:

• Does retention increase with increase in fun events?

• Is there a correlation between participation in fun events and
retention?
• Does engagement increase with more number of fun events?
• Does the same set of people participate in all programs? If
so, do they stay more?
• What about the engagement levels of people who lead the
programs? What about their retention levels?
• Is there a critical threshold of programs beyond which
participation drops?

Our analysis leads us to insight. These could be as follows:

Insight

a) There is little correlation between number of engagement

programs and engagement level.
b) There is a recency effect of big engagement programs. After
big programs, engagement increases for a short time.
c) There is little correlation between participating in fun
activities and engagement.
d) People who anchor cultural programs are highly engaged.
Attrition levels are less.
e) A minimum threshold of fun events is required. If it is zero
for a quarter, engagement drops marginally.

These and others could be insights. The insights not only validate
our assumptions but also help to frame strategy and take actions.

Action

1. Have not more than two fun events a month. Have one big
event every two months.
 Analytical Problem Solving 43

2. Identify and recognize the employees who volunteer and

anchor such events of their own interest. Consider such
behavior as well when identifying high potentials.
3. Create more opportunities for employee volunteering in
more areas including technology, CSR, referral programs,
etc. Validate whether ownership drives retention.

Chennai based Zoho corp has been able to develop a

competitive product without resorting to hiring engineers from
premier colleges. They have been successful in hiring for attitude
from high school pass students, training them in computers and
programming, and building productivity suites that are competitive
with the best companies like Google can offer (Shobna, 2015).
This is an example of a company using a contra strategy and being
able to pull it off by execution.

Strategy Validation

While the articulation can be different, every company has a talent

strategy. The strategy, in turn, is based on a set of assumptions.
It is not always that the assumption is based on insight from
analytics. Analytics may or may not validate the assumptions. That
should not come in the way of companies actually validating their
assumptions based on data (Figure 3.1).
Let us take a very simple illustration. Most companies say
that they create a “high performance culture.” On drilling down,
companies strive for high performance across different areas.
44 Winning on HR Analytics

Figure 3.1 Strategy Validation Framework

Source: Authors.

1. Hiring
a. Hire from the best colleges and universities.
b. Hire on the basis of consistently superior academic
performance.
c. Hire on the basis of a threshold performance on entrance
exams.
2. Performance management
a. Normalize employee performance.
b. Take action on the bottom 5%.
c. Align salary increases and rewards to performance
ratings.
d. Align promotions to performance ratings.
3. Development
a. Design competency framework based on superior
performers.
b. Select on the basis of competencies.
c. Train on the basis of competencies.

These are a few ways by which the company articulates a culture

that reinforces high performance. If we need to validate, the
questions should be at different levels. What are the questions?
 Analytical Problem Solving 45

1. In general, do people who scored more than 75% in college

perform better than those who scored 70%?
2. What does a plot of performance ratings versus percentage
marks look like?
3. Do employees from tier-1 colleges perform better than those
from tier-2 colleges?
4. What is the normal distribution for each college/company?
Does it show an abnormal pattern?
5. Does performance rating in one year predict the rating for
the next year?
6. Does an improvement in competencies lead to improvement
in workforce productivity?

Such questions throw up interesting answers that help validate

long held assumptions, some of which had become mythical.
Using a structured framework, companies have busted myths in
the following way:

• Sysco Corp., a S32 billion wholesale food distributor based

in Houston, found that its compensation system for drivers—
paying them by hours worked—did not provide as much value
to the organization as it could. “The model didn’t necessarily
provide better customer satisfaction or profitability,” says
Ken Carrig, executive vice president of administration and
head of HR. Instead, Sysco changed to a reward structure it
calls activity-based compensation. Drivers earn a base pay
that is supplemented with incentives for more deliveries,
fewer mistakes, and good safety records. Four metrics were
targeted: satisfaction level, retention, efficiency (delivering
more cases in less time), and delivery expense. Under the
new compensation structure, Sysco found that drivers were
not only more efficient, they were also more satisfied. The
company’s retention rates for drivers improved by 8%, and
expenses as a percent of sales went down. (Schneider, 2006)
• Using analytics, Google was able to identify the following
from a hiring perspective (Friedman, 2014):
a. Grade point scores have no correlation to on-the-
job performance. People who have a high grade point
46 Winning on HR Analytics

average (GPA) are good in taking tests in a structured

environment. This does not translate into real life.
b. Asking puzzles like “How many white cars are there in
Bangalore” actually do not result in better quality inter-
views! Rather, using a specific behavioral interviewing
technique delivers more consistent results. Even when
using such interviews, one gets more insight when asking
about a situation the candidate has handled, instead of a
hypothetical “how would you” question.
c. The company has even started hiring people with no
college qualification and it is working out well so far.
 4
Competing Through Workforce Analytics

O rganization structure is unique to every organization. Unlike

HR processes, it does not easily allow itself to be measured
for efficacy. However, improving a process produces benefits that
are much localized. On the other hand, any improvement using
analytics in the organization structure has greater leverage across
the organization. There are different facets to an organization:

• The number of levels/layers it has, which can also be called

an organization hierarchy,
• The number of functions/departments,
• Organization complexity. More the products/services
offered by the company across countries, greater the frag-
mentation of specialization and the need for people to
integrate across.
• Size: Larger companies often have more than one manager
per employee, through matrix organization structures.

Business Levers of Organization Structure

Textbooks have been written on how the organization structure
is one of the 7 Ss, with strategy, structure, style, etc. In a well-
structured organization, there are no superfluous roles or layers,
48 Winning on HR Analytics

and jobs are done at the right level of competency and expenditure.
For example,

• An organization looking to optimize its speed of decision

making reviews the number of layers it has in the structure
and reduces the unnecessary layers. Flat organizations are
supposed to have not more than 4–5 levels from the entry
level person to the CEO. Every additional layer adds to the
timeline for decision making.
• A well-structured organization typically is organized like a
pyramid, where the headcount decreases as responsibility
increases.
• In any organization, the individual contributors create
value while managers enhance the value by effective
administration. To be effective, an organization needs to
have the right ratio between the managers and the managed:
too high and the work outputs are not coordinated well
enough, too low and we have managers breathing down the
necks of employees.
• In matrix management systems, more than one manager vies
for an employee’s time. This in turn influences the social and
political system of the organization, impacting the culture.

Traditional Measures of Organization Structure

Traditionally, organizations have been using two measures for
assessing the health of their organization. Manager span of control
is the most popular measure. Quite simply put, it is the average
number of employees managed by a manager. We can arrive at this
by simply dividing the total headcount by the number of people
designated as managers. If a company has 100 employees and
10 managers, the manager span of control is 100/10 = 10.
Companies arrive at this ratio and set goals to improve it on an
ongoing basis. For example, if the ratio is 1:6.4 today, it should
become 1:7 tomorrow. This is based on the underlying assumption
that there is a good range for the ratio and one should strive to be
within that range.
 Competing Through Workforce Analytics 49

In the 1930s, V.A. Graicunas created a formula to identify a

desirable span of control. He estimated the span of control based on
relationships; every manager not only has to manage the direct rela-
tionship with each team member but also the relationships between
themselves and the group. When all these are added up, the manager
would find it difficult to handle more than six direct reports.
With time, attempts have been made to recalibrate this approach.
However, given the variety in operating models, one single formula
may not hold good, especially when we consider tasks to be as
important as relationships.
Obviously, any action cannot be taken on the basis of average
alone. The level of management and supervision required varies
on the basis of task complexity and team maturity. Consider an
assembly-line manufacturing. The supervisor has to maintain dis-
cipline and communicate the goals. The assembly line keeps them
occupied without any need for management. In such a situation,
one could even look at a management span of 1:20 or more.
On the other hand, in an R&D setup, there is a higher degree of
task complexity and the manager needs to work collaboratively
with the team. It would often be challenging for a manager to
aspire for a team size in excess of 4–5 associates.
Even within the same industry, the management span can vary.
In an ITES company, the span of control is very different for voice-
based processes and data analytics.
Studies have shown that the CEO span of control in the US has
increased from five direct reports in 1986 to 10 in 2006 (Neilson &
Wulf, 2012). Some seem to believe that 7–8 is a magic number of
optimal direct reports.
The manager span of control is a measure that every organization
needs to track, identify pockets of stretch as well as inefficiency,
and act accordingly. Span of control is also a requisite input to
workforce planning and helps a company arrive at the right number
of promotions.

Tooth-to-Tail Ratio
Tooth-to-tail ratio is used in the armed forces to arrive at staffing
levels. Combat personnel are the “tooth” and noncombatants like
50 Winning on HR Analytics

telecommunications, intelligence etc. are considered as “tail”. A

McKinsey study of 2012 estimates 26% as the tooth to tail ratio,
on an average 26% of the armed forces across different countries
are combatants. The figure ranges from 16% to 54%. (Gebicke &
Magid, 2012)
In corporates also, we have the frontline staff as well as the back-
office functions; we call them business functions and enabling
functions. Business functions are where the value is created for the
customer and delivered. Enablers assist the business functions and
support the requirements of the organization overall.

Typical business functions: Sales, manufacturing, customer support

Typical enabler functions: Finance, legal, and HR

A simple way of deriving this ratio is by generating the percentage

of business functions (sales + production + customer support) as
well as that of enabler functions on the overall headcount.
Fighting a battle calls for far greater preparation than
manufacturing and selling cars, for example. In comparison to a
battlefield, the parameters in business are more known and static.
As such, a 65 to 100 ratio would be an unimaginable luxury.
It is more likely to see 8%–12% range of headcount in enabler
functions, with the other 85%–90% headcount being engaged
in business functions. Cost cutting has heavily focused not just
on lesser proportion of staff in corporate functions, but has
gone forward with reducing the overall budget spent on such
functions. This, in turn, has been facilitated by the outsourcing
of a wide swathe of support functions, all the way from facility
management and catering to IT application and infrastructure
management. Offshoring, in turn, has helped to reduce such costs
further.
Often in MNC organizations, increase in headcount in enabler
functions is discouraged. Instead, third party outsourcing or
assignment-based consulting is encouraged. This approach helps
in spending just enough on deliverables instead of having a full-
time employee to manage and be managed.
 Competing Through Workforce Analytics 51

Much like the management span of control, the tooth-to-tail ratio

in an organization needs to be identified both at the headcount level
as well as the payroll level. An optimum organization needs to be
decided on the basis of business requirements, without providing
a buffer.
Often in HR, one finds a ratio based on headcount. We will have
one HR person for every X number of employees. X can vary from
100 to 500. This figure is a quick rule of the thumb number without
predictive validity. As an organization grows, one can follow two
approaches:

• Set a coverage ratio of 1:200 and then keep hiring HR

representatives to keep pace with growth.
• Set a target to improve the HR ratios even as the organization
grows. Go from 1:200 to 1:400.

HR ratio based on headcount also assumes that an HR person

is needed to perform several administrative functions and keep
up engagement. On the other hand, as an organization grows
automation improves helping every employee deliver more.
Increasing the number of HR partners has a lesser impact on
engagement than improving the quality of people management.
The role of HR is to ensure that managers do a great job of people
management. Then, HR partners migrate to a coaching and
consulting role from a pure relationship role. The HR headcount
diminishes but business outcomes get better.
Just following a standard ratio does not help us re-imagine the
function!

Becoming More Competitive Using

Organization Structure

Organization Shaping Through Pyramid Ratios

Ideally, every organization wants to be a like a pyramid. When
we are talking numbers, two dimensions suffice and it is really a
triangle, but the references have been to a pyramid.
52 Winning on HR Analytics

The logic is simple. It may be difficult to arrive at an exact

proportion of each level of employment. There are no basic ratios
like 1:3:9:36 and so on. However, companies believe that the
headcount at every higher level should be less than that at the lower
level.
Let us take the organization, whose example was shared
earlier. If we are to draw it into a pyramid, with length of the lines
equivalent to percentage headcount, it would look something like
this.

Role as a Role as a
percentage of percentage of
Role Role ratio lower role lowest role

Engineer 48% NA

Sr. Engineer 27% 1: 1.8 1:1.8

Lead 13% 1:2.1 1:3.7

Project Manager 6% 1:2.2 1:8

Manager 3% 1:2 1:16

Senior Manager 2% 1:1.5 1:24

Director 1% 1:2 1:48

In this table, the percentages indicate the percent headcount in

each role.

Role as a percentage of lower role

Percentage headcount in a role ^Senior Engineer h 27
= 2.1
Percentage headcount in the target role ^Lead h
= =
13

There are 2.1 senior engineers for every lead. This validates the
fact that the organization has a tapering pyramid.

Role as a percentage of lowest role

Percentage headcount in entry level ^engineer h 48
=8
Percentage headcount in the target role ^Project Manager h 6
= =
 Competing Through Workforce Analytics 53

For every project manager, there are eight engineers. This is a

healthy organization trend. If it were 2, then the company has far
too many project managers.
One would see that in this simplified example, the ratios
nearly double at each level. It can be equated to a progression of
2,4,8,16,32, and 64.
Each organization would have very different ratios based on
their stage of growth and internal mobility. It is not unusual for
manufacturing organizations to apply workforce ratios, primarily
to their white-collar staff.
Every company, after a point of time, sets up internal universities
to recruit fresh graduates from colleges to maintain their pyramid
ratios and as far as possible have talent grown from within. There
was a stage in the Indian IT industry when more than 60% of the
workforce was composed of entry-level engineers. This led to a
structure that looked like a triangle with its bottom stretched. This
is not an ideal structure either, as it puts too much workload on
managers to develop engineers.
In a healthy organization, it is natural that the role ratios between
a level and the one above it are usually more than 1. It shows that
in the organization, not everyone is promoted automatically.
Consider the following scenarios.

1. Role ratio is less than 1. Suppose there are more senior

engineers than engineers. Why does this happen? It could
be because engineers are actually trainees and so their
chances for doing billed work is less. It could also be that the
company has not hired at the entry level in the recent past,
as much as it used to. This could show a maturation process,
wherein the company has adequate headcount and is in the
process of improving the workforce productivity. Large
manufacturing organizations do not suffer from attrition and
so have stable employees at all levels. Campus hiring is just
to create a pipeline for future talent.
2. Role ratio is more than 3. This is the usual case at the level
of the CEO and his/her staff. There could be 10 SVPs but
54 Winning on HR Analytics

only one CEO. That is the last stop in a career. While that
is understandable, what about an organization that has a 4:1
ratio of lead to manager? This just means that the organization
is suddenly tapering off and the leads are going to take a long
time to become managers. Such a steep ratio would produce
challenges of job rotation and career development.

Organization Shaping Through Predictive Analysis

How does an organization stay on top of its pyramid ratios? One

obvious example is by growing talent internally and restricting
hiring strictly to entry levels. This way, there are always people
coming in through the ranks.
Second is to measure the role replacement ratios. Assume that a
company has an attrition of 10%. Then, it needs to replace that
extent of its headcount. Where there are internal successors
identified, people are moved into these roles.
External hiring is the second big source for replacement hiring.
It would make sense to divide replacement hiring into the following
categories:

• Replacement hired at a higher level,

• Replacement hired at the same level,
• Replacement hired at a lower level.

By default, when there is a vacancy, the job requisitions are

opened at the same level as the departing incumbent. It is often
ignored that the positioning was also due to the experience and
contribution of the incumbent. We need to review each position
and make sure that at least a quarter of positions annually are
actually replaced at a level less than the person leaving the job.
It would be fairly useful to review the % positions backfilled
at the same level and set a goal of at least 25% backfill at a lower
level.
Often, the usual representation of organization pyramids goes
something like this.
 Competing Through Workforce Analytics 55

This kind of a structure is the ideal. Each level has lesser people
than the level below that. However, this may nor may not be an
accurate representation of an organization.
Consider a multinational organization setting up operations in a
country. It will not have the liberty of hiring a lot of trainees from
campus at entry levels. If it is looking to grow fast, the company
would hire a lot of laterals at the higher band of specialist level as
well as at a team-lead level.
The company then could have a pyramid like this.

Let us examine this structure. Assume that the organization has

industry standard attrition rates. Over a period of time, employees
will expect growth. At the lower levels, they might expect to move
in 3 years, while the time period would be longer and even based
on vacancies only at senior levels.
The company then has to switch from a “default lateral hire”
to “default grow from within” strategy. This migration could take
3–4 years, failing which perceived growth prospects of employees
56 Winning on HR Analytics

would diminish. This would also switch the focus of hiring from
the specialist level to that of an individual contributor.

Year 1

Level Numbers Internal growth External hire

IC 30 30

Lead 50 10 40

Manager 20 5 15

Year 2

IC 60 60

Lead 70 40 30

Manager 30 15 15

Year 3

IC 100 100

Lead 80 60 20

Manager 40 32 8

As we see in this illustration, the company has grown from a

headcount of 100 to 220 over a 3-year period. However, it has
consciously changed the pyramid shape from one that was bulged
in the middle. Such a strategy also provides for a conscious and
optimal mix of internal promotions versus lateral hiring. Of course,
it is not an easy sell in the short term, as hiring managers always
seek ready replacements and do not give the impression of having
to allow talent to bloom. However, almost no great company was
built on a strategy of hiring ready talent from outside all the time.

Organization Shaping and Employee Growth

Employees would like their organization to provide consistent
career growth. Assume that an organization is in a growth industry
 Competing Through Workforce Analytics 57

and is growing as well as its competitors. Then, even if the company

has the ideal pyramid, it still needs to have a good distribution of
employees within the level.
For a long time, companies in India have been grappling with
the challenge of employee expectation of growth. It is felt that
employees would like an upward move, every 2–3 years. While
this has acquired the contours of a truism, how far is this possible?
Are all bands structured in a way that 80% employees can be
promoted within 3–4 years?
Employers are more liberal with career moves at the lower
rungs. With normal performance, an employee can hope to attain
3–4 promotions in their first 10 years. However, it is the rare person
who gets more than two or, at best, three promotions in the next
10. As vacancy-based growth starts at senior levels, backlogs get
transferred into the hierarchy. Expected tenure in band becomes
important.

Level Expected tenure in level

Engineer 3

Lead 4

Manager 5

Director 7

Given this reality, HR should share not just the career path
but also the time expected for each progression for average
performance and company growth rate. Then, for satisfying
employee aspirations, one also needs to review the population
distribution within each level.
This is an ideal situation. There are 20 managers eligible
for promotion, of which at least 15 get promoted, creating 15
vacancies. Then we have 30 aspirants from leads, of whom 15 can
be promoted. They can be backfilled by promoting from the 25
eligible engineers. This way employee growth can be cascaded
down the path.
58 Winning on HR Analytics

Level Tenure in level Numbers

Engineer 0–2 50

2–3 25

Lead 0–2 50

2–3 20

3–4 10

Manager 0–3 30

3–4 10

4+ 10

On the other hand, consider the situation where most of the

employees are in the lead level and the situation looks like this.

Level Tenure in level Numbers

Engineer 0–2 40

2–3 25

Lead 0–2 20

2–3 30

3–4 10

Manager 0–3 15

3–4 30

4+ 20

This organization is relatively top heavy. Of the 50 manager

aspirants, only 15 are promoted. For the 15 vacancies, there are
40 aspirants at the lead level and then there are 25 aspirants for
15 positions in the lead level.
Given that this shape has been created based on ad-hoc hiring, the
company will tend to lose more employees at the lower two levels,
as they have gained competencies of a higher level in their role and
they could easily move to the next level in other companies.
 Competing Through Workforce Analytics 59

Organization shaping then drives employee growth. When we

do not analytically review the staffing plans holistically, attrition
becomes an issue in the future.
For an organization to be more accurately represented, one
needs to show the headcount as well as the tenure in band in years.
This would also set the expectations correctly for employees and
they do not go away thinking that they can consistently expect to
be promoted every 2–3 years.

This representation also helps the employees to realize that there

is no automatic growth and different employees grow differently
on the basis of performance and assessed potential.

Look at Headcount in Offices

A company one of us had worked with had grown using
acquisitions. Each acquisition brought with it R&D centers
which were distributed across multiple locations. Over a period
of time, the company had offices operating out of 50 locations.
The company was incurring management expenses as well as
co-ordination expenses trying to manage all these locations. Then,
it took a decision to consolidate the headcount into 11 centers with
growth hubs identified in each continent.
60 Winning on HR Analytics

Often, companies end up in a sprawl of locations, each of which

gets created for reasons at different points in time. As always, 80/20
rule applies to this too. It would also help to review the headcount
at each location, identify the long tail and rationalize their number
on an annual basis.

Measuring the Softer Aspects of

Organization Structure
Organization structures look neat on the computer display.
However, in reality, they are filled by people who interact with
each other. Perception of the organization structure plays an
important role in enabling the success of the structure.
It is possible to measure some of these as a one-off exercise and
identify focus areas. For example, in a company with lots of matrix
reporting relationships, one can identify the number of employees
with matrix relationships and employees who are matrixed to
more than two managers. Both are indicators of organizational
complexity and need to be observed. However, identifying such
information takes time.
Employee surveys are most commonly used to capture all
perception feedbacks in companies.
A thoughtful organization design using analytics:

• Allows companies to be cost competitive,

• Increases employee motivation and reduces attrition by
providing career growth,
• Helps develop future-proof staffing plans.

Organization Demographics and

Succession Planning
Often countries are analyzed on their population distribution.
Europe and Japan, for instance, have ageing populations, while
India and China still have relatively younger population. India has
 Competing Through Workforce Analytics 61

the world’s youngest population which will remain so for the next
10–15 years. Analysts refer to this as the demographic dividend.
For a long time, Indian IT companies had an average age in the
mid to high 20s, representing their organization pyramid. Younger
the workforce, more the need to invest in retention as employees
are just getting started in their careers. Workforce becomes more
stable, once they get into the 30s.
This is what is attractively packaged as the “Gen X,” “Gen Y,”
millennials, etc. Adequate literature covers that. Let us check how
national populations are represented.
This is the graph for India. This graph not only shows the age-
wise distribution but also differences across gender. India has the
62 Winning on HR Analytics

perfect pyramid with percentage population reducing with age.

There would be enough youngsters to keep the economy running,
though it will be a big challenge to make sure all of them are
gainfully employed.
Let us look at Japan now.

Source: populationpyramid.net (accessed on July 28, 2016).

Japan has a different distribution of population. 40−44 and

65−69 are the age groups with most people. There could be more
people past the retirement age than the working population.
This puts a greater demand on social welfare. Just looking at the
demographics tells us that the nature of people problems in Japan
and India are very different.
 Competing Through Workforce Analytics 63

We have also come across some manufacturing organizations

where the average age is in the high 40s. These organizations also
have very low rates of attrition. What this means is that for several
senior roles, both the incumbent as well as likely successors are of
the same age and are likely to retire around the same time.
If we can visualize a similar representation in Indian companies,
the IT firms will be similar to those in India. The challenge is one of
engagement and employment. On the other hand, the visualization
for manufacturing companies will be similar to that of Japan—older
cohorts of retirement age. Succession planning in this case needs
to be derived also from the age distribution. Do we have enough
younger folks in the talent pipeline? Will we face a shortage of
people in 5–10 years’ time? How are we prepared for it?
Even in studies of gender diversity, we face the same issue.
Gender diversity is increasing with more women coming into the
workforce. At the same time, the representation of women decreases
with seniority. A representation similar to the ones shared highlights
these deficiencies in the manner that action can be systematically
executed.
 5
Acquiring High-quality Talent

T alent acquisition (TA) as recruitment is being increasingly

referred to as one of the biggest processes in talent manage-
ment. When the economy or a sector is growing very fast, TA
becomes a differentiator to business. Even during a relatively
steady growth, TA is key to bringing in people of the best quality
at the right cost. TA as a process has a lot of transactional elements
that make it more amenable to adapt an analytical approach. This
also hinders progress sometimes as the more meaningful analytics
dealing with quality need more coaxing and take a back seat. Here,
we will look at both aspects of TA.

Business Levers of Talent Acquisition

TA, more than any other process, reveals the talent strategy of the
company. A company could:

• Use a strategy of growing from within. Such a company

usually hires the most at entry levels and primarily from
college campuses. The company relies on a steady stream
of employees progressing through levels and occupying
leadership roles. Such a company rarely hires any lateral
talent from outside. It minimizes the cost of executive search
 Acquiring High-quality Talent 65

and the risk of a cultural fit, primary investment goes into the
development of talent.
• Use a strategy of need-based hiring. Companies in nascent
industries/fast-growth start-ups may not have the liberty
of growing all their headcount from within. They go for
need-based hiring and fill up positions across the board,
including senior management. Such companies balance their
investment into TA and talent development.

TA is also the only area in which companies compete with each

other in the HR domain. On business and engineering college
campuses, companies compete for talent. A better performance
there, should translate into a better performance at the business
front too.
In services companies, TA is also one of the biggest drivers
of cost. If a company is making a 20% increase in its headcount,
then the associated costs could be the equivalent of 3% payroll
cost. Finding an optimum mix of experience and sourcing becomes
important.
Often, in IT consulting organizations, one hears that:

• “Business is not the problem! Getting people is.”

• “We can win this deal, if only we have five people with this
skillset.”
• “We are not able to convert our offers in the US and business
is suffering.”
• “We will start this line of business as soon as we hire
someone who can drive it.”

Needless to say, getting the right person on time is very important

to keep businesses going.

Traditional Measures of Talent Acquisition

End to end talent acquisition process is depicted as follows. Often,
TA is measured on two axes: speed and effort.
66 Winning on HR Analytics

Speed

It would be ideal for the recruiting manager to fill in the position

as soon as it is thrown open. Imagine a vacancy opening up today
and getting filled tomorrow! That at least is the ideal as far as cycle
times are concerned.
In practice, hiring for each position goes through multiple steps.
First is the process of sourcing. For the given position, resumes of
eligible candidates need to be shortlisted. Then comes selection.
Shortlisted candidates undergo the selection process of interviews,
tests, and other assessment tools. This process takes time and can
often be the most time consuming of all the steps. Next comes
the process of making an offer to the selected candidate, verifying
her credentials in parallel (including background verification), and
having the person join.
There are two ways by which we measure this. For the hiring
manager, the clock starts ticking from the time a position becomes
 Acquiring High-quality Talent 67

vacant. It is his responsibility to get the work going. So, for him the
cycle time is the duration from the time a slot is opened to the time
it is filled. On the other hand, the TA team holds itself accountable
from the time the position is open to the time an offer is made.
After that, the joining time is influenced by the period of notice that
a new hire should give her current employer. Based on need, the
notice period can be bought out, but that is driven by the outcome
and not by an indication of the process efficacy.

Effort

Even till a decade back, some companies in technology used

to proudly mention the fact that they are receiving hundreds of
thousands of resumes. Even the ratios were tracked and there were
enviable numbers like only one in hundred applicants getting
selected.
Typically, recruiting teams used to work on the following
logic:

• How many offers we need to make for every join?

• How many interviews we need to do, for every offer?
• How many resumes we need to screen for every interview?

An inverted pyramid was formed, based on the following

metrics:

• Offer conversion rate (% join rate. If we make 100 offers,

how many join?)
• Interview conversion rate (If we do 100 interviews, how
many are selected?)
• No-show rate (How many turn up for interviews, if we call
100 people?)
• Resume conversion rate (If 100 resumes are shared, how
many are called for interview?)

Based on these numbers, the recruiting efforts were planned. Even

today, these measures lie at the center of recruiting planning, with
68 Winning on HR Analytics

companies using analytics to ensure there is better conversion at

each level. Let us just see what a difference of 10% in each step
does.

Process Company 1 Numbers Company 2 Numbers

Offer 80% 125 70% 142

conversion

Interview 70% 179 60% 238

conversion

No-show rate 20% 223 30% 340

Resume 60% 375 50% 680

conversion

In both cases, we end up with 100 new joins. However,

company 1 is able to achieve this with lesser effort. This in turn
means a greater process throughput. In the ideal world, to select
100 people, one should be able to just screen 150 profiles. Analytics
is already playing a role in the same, as we will see later.

Effectiveness Measures
Acquiring talent is not governed only by cycle times. There are
other dimensions as well. Cost is usually tracked by cost per hire.
How many rupees are spent for every new join? Identifying the
cost/hire is a great exercise, as it includes direct as well as indirect
costs. Usually,

Cost Cost (sourcing activity + technology + infrastructure + employee salaries)

=
hire Number of new joins

Cost of hiring drives several operational decisions in hiring:

• Balancing out between hiring channels: Job portals are

cost effective, but not sticky. On the other hand, referral
programs are more expensive but sticky in terms of higher
 Acquiring High-quality Talent 69

conversions. Cost per hire and conversion ratios help in

deciding what positions should be promoted for employee
referrals.
• Size of the TA team: Demand for recruitment varies with
business. Unlike other HR functions, it is difficult to keep
the team size constant. When demand is high, one needs
more people but when demand is less, one needs a smaller
team. TA teams typically have a more variable personnel
cost, often employing people on time bound contracts and
also by outsourcing the transactional activities.

Acquiring talent needs multiple channels: recruitment advertise-

ments, job portals, employee referrals, search firms as well as
social networks. How does one arrive at the right usage? Decision
making is based on the following indicative factors.
Assume that we need to hire 25 engineers next month. From the
table, it becomes clear that using job-boards and walk-ins should
be our primary approach; however, since the conversion rates are
low, this needs to be supplemented by employee referrals. On the
other hand, if we need to hire three regional managers in sales, a
walk-in interview will not help. Given it is a smaller pool, we also
need to ensure all offers are converted. Then, we rely on referrals
and use search firms.

Turn-
Set-up Recurring around
Channel cost cost time Scalability Conversion

Referral Nil Medium Medium Moderate High

Job board High Nil Fast High Moderate

Walk-in Low Nil Fast Moderate Moderate

Search Medium Nil Medium Low High

firm

These parameters are indicative. TA functions intuitively

use such a decision-making matrix when arriving at their mix;
especially in very large companies, employee referral schemes are
highly effective as they leverage the power of the network.
70 Winning on HR Analytics

Given that TA tends to use hybrid models of recruitment, it

is imperative that every full-time employee is able to deliver the
goals. As much as money, individual productivity is also important.
TA teams track it in two ways.
First is number of new joins/number of full-time recruiters.
Second is number of offers/number of full-time recruiters. Some
TA teams use the latter as they feel it is a better representation of
the efforts of the TA team; the former helps with measuring the
output.
Large organizations also segregate entry level and campus
hiring from lateral and mid-career hiring. Recruiter productivity is
a better indicator of performance for hiring laterals. On campuses,
it is the brand as well as the nature of employment that influences
the success rate. Companies use indicators like “Number of key
campuses on which the company was on day 0/day 1” to measure
the effectiveness of the campus hiring team.
Ability to hire within the range is another challenge that most
TA teams are grappling with, especially when it comes to lateral
hiring. Typically, a company has a salary range for every level.
Say for a team lead, it could be between `10 lakhs and `13 lakhs.
Ideally, this range should be spread equally across the performance
levels and tenure in the band. Suppose that a band has an experience
span of 4 years.
Then, the best performer with 3+ years of experience should
be getting 13 lakhs, while a person with highest rating but 1 year
experience may get 11.5 lakhs. There are underlying complexities,
but the more experienced and better performing a person is, the
higher is his positioning on the band.
A company would ideally like all new hires to come within this
salary range with similar rules. There may be a need to hire a few
people even outside this range.
Let us illustrate this as follows.
We are sharing two simple distributions. Newhire 2 has more
people in the lower third and lesser people in the upper third. On
the other hand, the situation is reversed in Newhire 1. It is to the
skill of the TA sourcing team and the brand of the company if
the company is able to maintain the same graph for new joins as for
 Acquiring High-quality Talent 71

existing employees. Go lower and you risk a lower conversion, go

higher and you risk employee unrest!

Emerging Measures of Talent Acquisition

We had earlier spoken about cycle time and cost of hiring. Under
normal conditions, one would try to balance out the two.
Let us consider that we have 25 positions to be filled in regional
sales. The company is losing market share fast. It is imperative that
the new sales folks will need at least 2 months to become produc-
tive. Unless they join in a month, the company will struggle to meet
its revenue goals for the year. It may need TA to use consultants
more and do sourcing across the country. The company may even
need to buy out notice periods. The cost of hiring will increase in
the short term.
In this situation, will TA be focused on the cost of hiring or on
the value of adding new hires in the quickest possible time and
gaining sales?
Increasingly, experts are beginning to point out that while cost/
hire is a useful measure, it should not be the primary measure. Cost
should also be seen together with criticality of the position and the
value added by that position.
72 Winning on HR Analytics

Opportunity Cost of Cycle Time

When backfilling a role, there is another important dimension that
is not sufficiently articulated. It is often assumed that a person
leaving a job, under normal circumstances will be performing
at near full ability. However, once the person decides to leave,
his workload is distributed and the efficacy comes down by half
or more before the day of separation. On the other, a new hire,
however well qualified he may be, starts at zero. It takes time for
a person to understand the organization, operating model, etc. It
might take 3–6 months before the replacement starts operating at
near optimal levels.
This drives the business case for retention. Let us consider a
simplified hypothetical example. An employee is paid `50,000
a month and at the least delivers a value equivalent to the same
amount. April goes on fine. But by the end of April he decides to
move on. Organization triggers hiring actions but at the same time
reassigns work. If the employee has 2 months’ notice period, then
he may be delivering at half his effectiveness in the first month and
nearly at zero in the second month. Assume the replacement joins
on 1st June. Using the same model, it would take 2 months for him
to be at an optimal delivery level.
Apparently, things are fine and the position has been occupied
continuously. On the other hand, the company had incurred a
cost of `300,000 in this period, while the returns have only been
`150,000.

Month Cost Value

April 50,000 50,000

May 50,000 25,000

June 50,000 0

July 50,000 0

August 50,000 25,000

September 50,000 50,000

 Acquiring High-quality Talent 73

Does this mean that it is okay to let the job vacant, without
incurring a cost? Not really; there is opportunity cost of business
not fulfilled or customers not handled as well as they should. This
is felt by the hiring manager and is transferred onto the TA team.
The following graph illustrates the same.

Validity of Hiring Specifications

What happens when a company is flooded with resumes and
needs to select 1 out of 50 or one out of 100? There needs to be
a standard set for shortlisting and processing these applications.
Given the large numbers, any standard that is set needs to be simple
and appear fair. Such information should also be available for all
candidates.
What is so easily available? Only college education as
performance in college is universally applicable. So, companies
start by drawing the guidelines on the basis of the following:

• Academic qualification (diploma/graduate/4-years graduate/

post-graduate/PhD, etc.),
• Type of graduation attended (arts/sciences/technology/
management),
74 Winning on HR Analytics

• Quality of institute (premier/well-known/emerging/not

known),
• Performance in education (first class/70%/consistent first
class, etc.).

This makes sense in a country like India, where often a professional

education qualification is seen as the preferred choice for good
students. Most good students go in for an engineering graduation.
The better students clear tests to get into premier institutes such as
the IITs and NITs. Even within these institutions, there is a pecking
order in terms of specialization. For a computer engineering aspir-
ant, the choices would be computer science, IT, electronics and
communication, electrical and electronics, and so on.
A company hiring for a certain number of engineers then follows
an algorithm that looks like this:

• Engineering graduate or no?

◦ If engineering graduate, premier college or no?
▪ If premier college, first class or no?
▫ If first class, consistent all through or no?

Based on such rules, candidates are shortlisted. The criteria are

objective and can be shared transparently. However, this also
creates a scarcity-based talent model. Students who tick all the
boxes are in heavy demand from multiple companies. At the base,
students who tick off just one or two boxes end up competing
with a lot of others and can even end college without being
employed.
It is not unusual to hear from companies that there is a shortage
of good talent. In a country of 1.2 billion people, this is surprising.
So, what is going wrong?
While the criteria are objective and measurable, they do not
accurately measure what is needed to perform well on the job. Does
an IIT qualification automatically mean that the best programers
come out of there, to the exception of every other pedigree? Do the
best engineers always come from the best schools?
Researchers have been analyzing the correlation between
on-the-job performance and academic qualification. More than one
 Acquiring High-quality Talent 75

study has found that there is little correlation between the academic
background of a person and their on-the-job performance. In the
words of the Google head of People Operations “noted that Google
had determined that ‘G.P.A.’s are worthless as a criterion for hiring,
and test scores are worthless.... We found that they don’t predict
anything.” He also noted that the “proportion of people without
any college education at Google has increased over time”—now
as high as 14% on some teams.
That is not to say a college education is not needed or a high
GPA is bad. It is just that other things being equal, academic
performance or even performance on a hiring test does not indicate
future performance.
Future performance is predicted by competencies a prospective
employee possesses. Google talks about humility, ownership,
learnability, and emergent leadership. These can be assessed only
through structured assessment of competencies.
It is often amusing that most companies in India are faced with a
talent shortage and there is high competition for good programming
talent, for example. How can a country of 1.2 billion have a talent
shortage?
That happens because most companies are focused on the tip
of the iceberg—people working in successful companies with a
pedigreed degree. On the other hand, if the problem is redefined
and they start looking at alternative talent pools and select on the
basis of competencies and train, the challenges will be mitigated.
That in turn needs use of analytics to identify what competencies
matter and how to develop them.

Importance of Quality of Hire

TA teams world over hold targets for meeting numbers, service
levels and costs. However, not often does the quality of hiring gets
reviewed or assessed. How does one create a TA function that
delivers quality workforce?
Some companies have tried to put a quality of hiring measure.
Let us go through different measures first.
76 Winning on HR Analytics

As discussed earlier, one can start by looking at the pedigree of

the candidates:

• What percentage of new hires has a professional

qualification?
• What percentage of new hires is from named competitors?

In the absence of everything else, these measures act as a surrogate

for quality. As discussed, competency-based hiring is a better pre-
dictor of on-the-job performance. However, what is the first step to
competency-based hiring? Enough selection tools should be created
for hiring for competencies and interviewers trained in that:

1. What percentage of roles has competency-based selection

tools?
2. What percentage of critical roles includes validated
assessment tools?
3. What percentage of interviewers has been trained in
competency-based selection?

These measures ensure that the checkpoints are aligned with

competencies needed for quality hiring.
Once the new candidates are on board, it is expected that they
reach an optimum level of performance within 3–6 months.
In companies, this is given as the probation period, after which
the employee is confirmed in service. In case the inputs are of
inconsistent quality, there would be variations in the confirmation
rate of new employees. This can again be tracked as:

4. Percentage of new hires whose confirmation is delayed.

5. Percentage new hires that are let go within the 1st year on
performance grounds.

After a year, the new employee has integrated with the company
and is being assessed for performance. An analysis of performance
review information, in turn, helps us create a quality measure:

6. What percentage of new hires having less than 6 months

experience gets an “above average” rating.
 Acquiring High-quality Talent 77

7. What percentage of new hires having less than 6 months

experience gets an “average” rating.
8. What the performance rating distribution is of new hires
versus the distribution for existing employees.
9. New hire retention levels.
From the above-mentioned list, it seems like the best way of
measuring quality of hires is to use a combination of measures. It
could be a challenge to create a composite measure unless we assign
relative weightages to the components of the quality measures.

Quality of Hiring is More Important than

Plain Cost of Hiring

Let us take two companies and assume they are hiring for similar
positions. One of them spends just `25,000 per hire, while the
other spends `35,000 as it wants to get the fit right.

Year 1 Year 2
Cost of replacement replacement
Company hiring cost cost Total cost

Company 1 25,00,000 8,75,000 9,25,000 43,00,000

Company 2 35,00,000 5,25,000 2,40,000 42,65,000

Let us look at the cost for 100 new hires.

Company 1 incurs a cost of `25 lakhs for 100 hires while
company 2 incurs a cost of `35 lakhs. The finance and HR
organizations are happy in company 1 having saved `15 lakhs.
Let us look at the second year. Company 1 has focused on cost
but not the profile of workforce that would help in retention. To
some extent, the first year costs for company 2 were higher as they
also invested in recruiting instruments based on analytics. Of the
100 people who joined, company 2 has an attrition of 15% while
company 1 has an attrition of 35%. In the second year, company 1
has an attrition of 25% while that for the same group of people in
company 2 has come down to 5%.
78 Winning on HR Analytics

Assume that the costs remain same and the companies refill
all vacancies. Let us also assume that attrition affects backfills
in the same ratio. (Of 35 people hired as replacements, 35% quit in
the second year).
Let us look at the costs now. At the end of year 1, company 1
has saved 40% when compared to company 2. But after 2 years,
the running costs of company 1 are more and even though the cost
of one-time hiring was higher for company 2, the overall costs are
actually less.
This is not to say that a higher quality hiring process should be
more expensive all the time. However, it is important that hiring
costs are blended with other workforce quality measures to arrive
at qualitatively better decisions.

Talent Acquisition for Predictable Joining

and Performance
Let us go back to our original example of a company needing to
select 100 new hires. Let us take the more productive company, but
expand to include the hidden costs of recruitment. For simplicity’s
sake, let us just focus on the time managers spend:

Time
Additional spent by
Process Company 1 Numbers effort managers

Offer 80% 125

conversion

Interview 70% 179 54 108 hours

conversion (179–125)

No-show 20% 223 152 25 hours

rate (375–223)

Resume 60% 375

conversion
 Acquiring High-quality Talent 79

Let us review the numbers. The managers screened 152 resumes

more than required, resulting in 25 hours of additional work
(assuming 10 minutes per resume screening to shortlist). They also
interviewed 54 more candidates, so as to provide a buffer for the
final joining numbers. At 1 hour per interview for two managers,
this is another 108 hour additional effort.
In this entire process of using a hiring pyramid, 133 hours of
extra management time has been spent. In other words, for each
join, the hiring manager spends at least an hour on wasted activity.
16 person days of management time, is at least `100,000. In other
words, the hidden cost of hiring efficiencies adds `100 per offer
in hiring costs!
This is just a very conservative estimate. A manager has many
priorities and spending such time takes attention away from other
priorities. This time is immaterial when one is hiring for senior
positions or roles with a unique skillset. On the other hand, TA
incurs such extra efforts even for fairly generic roles. Analytics
helps us question these long-held assumptions. Can a company use
analytics to identify the exact fit people, who would be willing to
join, much like sabermetrics for TA?
It then becomes important to create an ideal state. If I need to
have 100 joins, can we interview just 120 people? Can we get that
120 from just 200 resumes? Instead of a taken-for-granted pyramid,
can we do with a tapering rectangle?
Analytics companies are helping solve this exact problem.
Technologies used are:

• Context-based search. Resumes are written freestyle. Earlier

search technologies used keyword search to identify skills.
But that is not so useful any more. The new algorithms read
the profile of joins and identify what is the characteristic of
new joins in terms of
◦ Nature of company,
◦ Experience level,
◦ Location,
◦ Technology,
◦ Compensation.
80 Winning on HR Analytics

  Then they identify the parameters that distinguish the

joins. For example, they can plot a zone of join based on
company, location, and joining salary; then they can target
that particular profile to get more conversions with the same
effort.
• Companies are also using competencies to move to hire for
assured performance. We might make big investments and
hire people of the same quality. The new talent pool with
luck would be as competent as the existing pool. Ideally,
it should be better. Companies analyze the competency
profiles of existing employees and then try to hire candidates
with similar profiles from the talent pool.

This increases the design time for HR and training time for
managers—but ideally, if the results are not only a more productive
process but also a higher level of performance, why not?

Company Company Company Company

S. No. 1 2 S. No. 1 2

1. 65 85 10. 80 95

2. 70 90 11. 105 85

3. 75 95 12. 75 90

4. 90 95 13. 80 95

5. 105 102 14. 110 100

6. 120 99 15. 130 95

7. 140 103 16. 90 90

8. 90 95 17. 120 101

9. 65 91 18. 75 95
 Acquiring High-quality Talent 81

Measuring and Improving Process Capability

Manufacturing processes need to conform to high standards of
reliability. Usually there are two measures to evaluate the per-
formance of a process. One is process control. Process control is
the “voice of the process” that is derived from how consistent the
process has been performing over a period of time and indicates
the consistency of output. Process capability, on the other hand, is
a measure of goodness of the process comparing the “voice of the
process” with the “voice of the customer”.
Applying to TA, process control gives a view of the number
of joins/offers the process delivers on a consistent basis. Process
capability will give an indication of how well it is able to meet the
customer needs.
Let us take two different scenarios. For the sake of brevity, let
us take performance over 18 months of a recruitment function. Let
us consider that the monthly demand is 100 joins in both cases. The
actual performance of the two functions is as below.

Company SD Process control Process capability

Company 1 22.41 0.3 0.2

Company 2 6.36 1.3 1.0

For simplicity’s sake let us assume that the business has a

tolerance of 20, even though it is a wide range! The customer has
a target of 100 per month and is okay as long as the joins are in the
80–120 range.
The formulae for process control index (Cp) and process
capability index (CPk) are as follows:
(USL - LSL)
Cp =
(6 # SD)

For the purposes of this exercise, the upper specification level

(USL) is 120 and the lower specification level (LSL) is 80.
CPK is least of (USL - mean) /(3 # SD) or (mean - LSL) / (3 # SD)
82 Winning on HR Analytics

Standard deviation (SD) for company 1 is 22.41 while that for

company 2 is 6.36.
The values of Cp and CPk are as follows:

What are the takeaways from this?

1. On an average, company 1 gets 93 joins a month, while

company 2 gets 96. Not a big difference.
 Acquiring High-quality Talent 83

2. However, the process control index Cp for company 2 is

1.3. This indicates that the process is good and the amount
of reviews can be reduced. On the other hand, the Cp for
company 1 is 0.3, indicating an unpredictable process.
3. Let us look at CPk now. The process in company 1 is more
within the desired range, but is there any skew? The CPk
for company 1 is 0.2. This indicates that the process is not
meeting customer expectations. Not only is the process not
meeting expectations, but also for any given month, it can
be higher or lower than the expectations. Company 2 has
a CPk of 1. While this is not ideal, in comparison with
Company 1, they deliver more within the range and the
numbers also reveal that they consistently deliver a little
below the goal of 100.
  For a goal of 100, Company 1 may deliver between 70
and 130, while Company 2 would deliver around 95. This is
a more predictable process, in need of fine-tuning. A visual
representation of the same is given below and it is obvious
that the process is far more reliable in company 2.

One could argue that measuring TA processes with the rigor of

six sigma adds more challenge. However, usage of concepts like
ranges, limits as well as SD provides a lot of clarity, especially
when the nature of demand is uniform.
Articulating using process language also helps the business
to take into consideration the variabilities so that their business
planning can be more grounded.
 6
Results-oriented Talent Development

S ince the publication War for Talent by McKinsey in 1997

(Michaels, 1997), finding the right talent has been an ever
growing challenge for the companies. And the process of finding
the right talent is not cheap. There have been various studies
to estimate how much a new hire cost to the company from the
joining date to the date of being productive, and the rule of thumb
being that based on the level of hire, cost can range from 1.5x to
3x of the salary for a fully productive employee. Imagine the total
cost for hiring 1,000, 5,000, or 10,000 employees per annum. Now
here is the challenge. A company may be able to find the talent and
bring them on board by spending huge amounts of money but any
leakage of the talent will hit company manifold in costs. Hence,
after hiring, the next big challenge for any company is to retain
the talent as the replacement cost per employee will be even much
higher than the hiring cost!
Talent development works as a double-edged sword: first, by
developing employee expertise or skills to help the company deploy
trained employee on higher positions and second by retaining
the employee for longer periods as employees start appreciating
benefits of longevity than short tenures. Following are some of the
reasons why companies invest in talent development:

• Skill development,
• Retention,
 Results-oriented Talent Development 85

• Ensure availability of roust talent pipeline,

• Enhance employee engagement,
• Increase job performance,
• Foster continuous change,
• Employer branding.

There are various ways of talent development in any organization

and these are often categorized as formal and informal methods.
Classroom-based development is formal, which is complemented
by other methods called informal methods. Broad list of informal
methods includes:

• E training or virtual learning,

• Mentoring,
• Coaching,
• On-the-job training/learning,
• Informal or social learning,
• Self-directed learning,
• Job rotations,
• Project assignments,
• Attending conferences and seminars,
• Tuition reimbursement.

All of these talent development methods involve cost—both

in terms of direct cost spent on these activities and indirect cost
based on employee time spent on these activities. Globally, train-
ing spend per employee (2014) ranges from USD 1,200 to USD
2,000 and average training hours per employee range from 36 to
42 depending upon the size of the company and industry type. In
India, the average employee spend in year 2014 was USD 375 and
the average number of training hours ranged from 40 to 49 hours
per employee. Across the globe, companies in the year 2015 spent
USD 355.6 billion on employee training (Training Industry Inc.,
2016) and in India it is estimated by various sources to be around
USD 26 billion.
Above cost figures show that there is a significant spend on
training by companies. As a rule of thumb, cost of training in any
organization ranges between 2% and 3.5% of the payroll cost.
86 Winning on HR Analytics

In companies where the payroll cost is 60% of the total revenue

like services companies, training spend becomes a significant
cost, forcing the C-level to ask questions on what is the return on
investment (RoI) on the training budget. And if the data shown
to the C-level for demonstrating RoI of training spend are not
convincing, then the training budget is the first casualty when any
company or economy is not performing well.
Measuring RoI of talent development initiatives has always
been a challenge, but it is seen as a priority by many organizations.
A study by Palmer (2010) found that only 8% of the companies
actually evaluate return on talent development initiatives. So how
can a company measure value of investment in a talent development
initiative? The earliest known and popular model has been given
by Donald Kirkpatrick’s (1959a, 1959b, 1960a, 1960b) four-level
learning evaluations model published in a series of four articles in
the Journal of American Society of Training Directors. The four
levels of the model are:

• Level 1: Reaction—captures reaction of participants

immediately after the intervention/program, is over, mainly
in the form of smiley sheets regarding participants’ views on
logistics of program, facilitator, content, etc.
• Level 2: Learning—captures participants’ response on
knowledge or skill increase after attending the program.
• Level 3: Behavior—this level measures the application of
learning by the participant while on the job after attending
the program.
• Level 4: Results—here focus is to capture change in the
performance levels of an employee after attending the
program.

These four levels still fall short actually connecting investment

with business level outcomes such as profitability and revenue
growth. Jack Philips (1996) added the fifth level as RoI.

• Level 5: RoI—to measure benefits over the cost invested for

a program or initiative.
 Results-oriented Talent Development 87

Later on, two more levels were added—Level 0 as the base level
and Level 6 as the optimization level by Pease, Bradford, and
Walker in 2014, making the total number of levels 7. Briefly, these
levels mean:

• Level 0: Base level—this level captures basic data like

utilization rate of each program/initiative meaning how
many participants attended the program against the available
seats.
• Level 6: Optimization level—this level attempts to show
the actual impact of talent development investments on
business outcomes such as revenues growth, profitability,
productivity, retention, etc., by connecting the dots, and also
goes a step further in highlighting where talent development
initiatives are working or not, thus giving choice to company
for optimizing investments for higher returns.

So all the seven levels can be shown in a model linking business

impact and analytics value-add of each level as shown in
Figure 6.1.
Let us illustrate the above-mentioned seven levels with the help
of an example so that talent development practitioners can apply
the same at their workplace. ABC Company hires 300 engineers
with 3–4 years’ experience in 2014. After 2 months of joining,
150 engineers are put through a two weeks’ training program.
ABC Company can use the seven levels of talent development
measurement as follows:

1. Level 0: This level captures basic data; in this case, 150

persons were nominated and 140 participated successfully
for two weeks.
2. Level 1: This level simply requires administering feedback
forms to those who attend and collect the feedback related to
facilitator, logistics, etc., and analyzing feedback form data
and then using visualization techniques such as histogram,
graphs, pie charts, etc., to present the feedback data to
stakeholders.
88 Winning on HR Analytics

Figure 6.1 Data-based Talent Development Measurement Value Chain

Source: Authors.

3. Level 2: This level is related to learning gain by the

participant. This can be measured typically by conducting
a quiz-type test within one month to measure knowledge
retention by the participant.
4. Level 3: This level requires measuring changes in participant
behavior and for this a structured form needs to be used to
 Results-oriented Talent Development 89

capture the behavior changes displayed by the participant

who has undergone training. The form has to be filled by the
manager of the participant with the form having columns like
the list of behavior changes intended, pre-training behavior
displayed, and post-training behavior displayed, using the
Likert scale to measure responses.
5. Level 4: This level concerns the results level and requires
capturing either monthly or quarterly on KPI indicators
mapped to the competencies/skills intended to be improved
by the program attended by the participant. If the program
focused on handling difficult customers or presentation
skills, etc., then improvements in these as reported by direct
manager can be used as a fair measure to capture the benefit
of behavior change (level 3) leading to right outcomes, that is,
improved customer handling of the quality of presentation.
6. Level 5: This level requires measuring RoI of the program.
The best method to measure this is by using the A/B group
testing or split testing technique. In this technique, the group
which underwent training is called the controlled group (A)
and the group which did not receive training is called the
non-controlled group (B). For example, if the objective of
the training was to improve dealer satisfaction or increase
product sales by the sales team and if data show increase
in dealer satisfaction levels or product sales figures for
Group A as compared to be Group B, then it is fair to
conclude that training has contributed to business outcomes
(dealer satisfaction/increased sales) like revenue increase.
Ideally, if resources permit, any organization should use A/B
testing for levels 3 and 4 also.
7. Level 6: This level deals with the optimization and predictive
effectiveness of the training to the participants. This level
helps in measuring the ultimate utility level of a training
program to the participant and organization. Let us use an
example of ABC Company explained earlier. Out of 300
engineers, 140 attended the program. Let us say that 1 year
after attending the program, 80 out of 140 got promoted to
the next level. How the Bayesian theorem can be used to
find out the predictive effectiveness of training for those
90 Winning on HR Analytics

Table 6.1 Applying Bayesian Theorem to Training Impact

Training attended Promoted Not promoted Total

Yes 80 60 140
(57.14%) (42.86%) (100%)

No 55 105 160
(34.37%) (65.63%) (100%)

Total 135 165 300

(45%) (55%) (100%)

Note: Row percentages in parentheses.

who attended the training program? Here is how this is done

with help of the Bayesian theorem by putting data in a matrix
form (Table 6.1).

Predictive Effectiveness of Training

Odds of being promoted after training
= # 100
Odds of being promoted without Training

where
Probability of being promoted after Training (p)
Odds of being promoted after training
= # 100
Probability of not being promoted after attending Training (1 - p)
57.14
= = 1.33 or 133%
42.86

Probability of being promoted without Training (p)

Odds of being promoted without training
= # 100
Probability of not being promoted without Training (1 – p)
34.37
= = 0.5437 or 52.37%
65.63

133
Predictive Effectiveness of training = = 2.54 or 254%
52.37
 Results-oriented Talent Development 91

Interpretation: Predictive effectiveness of training calculation

shows that chances of being promoted are 254% higher for those
who attended training than those who did not attend training,
everything else being equal.
Hence, a benefit of attending training improves chances for the
individual to get promoted and for the organization it indicates that
the use of training leads to better outcomes in future.

Measuring Return on Investments on

Talent Development Initiatives
As technology is advancing fast and organizations are witnessing
digital transformation, it is becoming easier and cheaper to collect
and store vast amounts of data related to talent development
initiatives and connect it with various business levels and other
outcomes relevant for the organization. As the talent development
initiatives can impact vast parameters of people management,
typically organizations across the globe use the following metrics
to gauge the impact of talent development initiatives:

• Employee engagement,
• Employee retention,
• Business-like productivity, reduction of costs,
• Customer satisfaction,
• Reduction in failure rates,
• Talent pipeline depth,
• Leadership performance,
• Innovation,
• Teamwork.

Measuring all of these parameters is not easy as some of the

development initiatives pose measurement challenges. For
example, measuring informal or social learning or self-directed
learning is not amenable to easy measurement. At the same time,
if development has become strategic and shows business outcome
impact, then it has to move from tactical skill type focus to
connecting development initiatives to business performance. For
92 Winning on HR Analytics

the sake of making measurement easy and meaningful, classifying

outcomes into different levels as follows may help:

1. Organizational level,
2. Individual level,
3. Program/Intervention level.

Another approach for classifying metrics and measures is to align

them to the needs of business leaders so that investment connects
to outcomes and business performance is demonstrated. Using this
as a guiding principle, we can categorize outcome-based metrics
under three broad categories as shown in Figure 6.2:

1. Organizational capacity-related: This category captures

metrics related to building and growing talent depth in the
organization.

Figure 6.2 Typology for Talent Development Measurement Levels

Source: Authors.
 Results-oriented Talent Development 93

2. Organizational capability-related: This category covers

metrics capturing skill or competency development.
3. Organizational performance-related: This category captures
metrics related to business growth.

Identification of measures or metrics1 that will be relevant for

each level and those which can be connected to business outcomes
will further provide clarity on what to measure and how to measure.
Any measure must satisfy some criterion to be qualified as a metric
fit for measurement. Here is a list of those characteristics:

1. It should be fairly usable by all the types of organizations

though each organization may need specific measures for its
context.
2. It should be easy to understand and deploy.
3. Data should be easily available or identifiable with some
effort.
4. It should make use of current data as well as should be
flexible enough to incorporate new data.
5. It should provide insights into both leading and lagging
indicators.
6. It should have link or should be connected to performance as
otherwise it does not make business sense.
7. Metric should be credible with stakeholders.

Right Metrics and Measures for Strategic Alignment

What will be key metrics and measures for talent development
initiatives which will combine all the characteristics of good
metrics and measures listed above? Practicality demands that the

1
 Metrics and measures are used interchangeably in the business context though
there are subtle differences. Metric is a derivative of measure. For example,
the number of training hours offered in year and the number of training hours
utilized in each program are measures. Metric for this will be the utilization
rate which is derived by dividing the total number of training hours by the
actual training hours used/attended.
94 Winning on HR Analytics

number of such metrics and measures be minimum, otherwise a

long list will become overwhelming for any talent development
practitioner to track and meaningless for stakeholders from the
sense-making point of view. Before covering examples of metrics
under each category, it will be useful to provide a linkage between
all categories and business outcomes with the help of a model or
framework as shown in Figure 6.3.
A map linking various talent development interventions where
typically any company makes investments and their linkage to
business results are shown in Figure 6.4. In this figure, connection
between talent development programs, indicators, metrics, and
business results including financial performance is presented
in a simple manner. However, in each company such flow
connecting various elements will vary depending upon the type
of talent development programs in use and business outcome to
be measured. It will also vary due to a company’s priority at that
particular point of time to measure the impact of investment in a
particular program in use. Another cause of variation will be the
type of data available and the ease of measuring such data. Based
on the context and feasibility, each company can broadly draw
a talent investment–business outcome linkage map as shown in
Figure 6.4.
According to Lavoie (2014), business leaders across the globe
are increasingly looking for data to make talent development
investment decisions. However, a CEO survey done by PwC in
2014 shows that 80% CEOs want data for talent development
investments but only a small percentage is able to receive it
(Saratoga, 2015). Another study by Vickers (2010) i4cp found that
high-performing companies more often measure talent investments
and some top measures used are as follows:

• Financial performance,
• Talent pipeline,
• Cost reduction,
• Employee engagement,
• Leadership success.
 Results-oriented Talent Development 95

Figure 6.3 Metrics for Measuring Impact of Talent Development

Investments

6XSHUYLVRU0DQDJHU(YDOXDWLRQ

&DSDELOLW\
3URGXFWLYLW\
5HODWHG

3HUIRUPDQFH4XDOLW\

,QWHUQDO+LUH5DWH±3URPRWLRQ
7UDQVIHU
0HDVXULQJ%XVLQHVV
&DSDFLW\
,PSDFWRI7DOHQW 6XFFHVVLRQ'HSWK
5HODWHG
'HYHORSPHQW
0RELOLW\5DWH±1RUPDO+L3R

&OLHQW6DWLVIDFWLRQ

&OLHQW5HIHUUDOV

2UJDQL]DWLRQ &OLHQW5HQHZDO&OLHQW:DOOHW
5HODWHG 6KDUH ,QFUHDVH

3URSRVDO:LQ5DWH

5HFRPSHWH:LQ5DWH

(PSOR\HH (QJDJHPHQW

(PSOR\HH
(PSOR\HH 7XUQRYHU
5HODWHG

5HGXQGDQF\5DWH

Source: Authors.
Figure 6.4 Talent Investments–Metrics–Business Outcomes

Source: Authors.
 Results-oriented Talent Development 97

Here are some industry examples of talent development investment

measurements:

1. Sun Microsystems used predictive analytics to measure the

impact of a mentoring program on financial performance.
Sun did a two-phase study of comparing a control group with
an experimental group for a mentoring program. Results of
the analysis showed that $6.7 million were saved through
improved retention, while the total cost of the mentoring
program was $1.1 million (Sun Microsystems, 2014).
2. ConAgra Foods created a Foundations of Leadership (FoL)
Program to align leadership across the company. The
RoI of the program was studied by third party—Bellevue
University’s Human Capital Lab using a control group
approach. The study analyzed the performance of 600
trained and 1,600 untrained supervisors and found that the
program improved the productivity and reduced employee
turnover. The study further found that those who went
through the training improved their chances of promotion
by 2x (ConAgra Foods, 2010).
3. GoogleEDU, the employee training division of Google,
makes use of data in aligning training to business strategy.
For example, it uses data to offer more programs to certain
employees than others and this is based on data collected
from manager evaluation of the employee performance and
training transfer, suggestions by team members, career stage
of the employee, and even location. Google also uses data to
create training redundancy by eliminating training programs
which are no longer working, and offer more of those
programs which are delivering better employee performance
(GoogleEDU, 2012).
 7
Talent Engagement and Retention

T alent engagement is an idea whose time has come. In fact,

some argue there is far too much focus on just engaging the
employees at the cost of enabling them to align their performance
to that of the company. Just how much engagement is required?
Engagement has one of the tent poles of HR analytics. Most
technology companies rely on human capital. Bill Gates was
supposed to have once remarked, “The market capitalization of
Microsoft is zero at the end of the day. It does not come back up
unless all of them return to work the next day.” Obviously there was
some maturity in tracking attrition, analyzing it, and identifying
patterns in the same even before analytics became a happening
area. Tools and platforms have been created that help in integrating
existing information and create a degree of predictability.

Business Levers of Employee Engagement

Initially, companies counted only one measure to track employee
satisfaction. It was felt that satisfaction would lead to motivation
which would lead to performance. Turnover rate was taken as the
measure for satisfaction. If the employee turnover rate is less, then
they are more satisfied.
Workers in factories hardly quit. However, there are also
cases of workplace disruptions including strikes that shut down
 Talent Engagement and Retention 99

plants. Does it mean turnover alone is not enough to measure

satisfaction?
So HR started adapting tools from market research. Like
customers, employees were also surveyed to find out how satisfied
they are with the company. Normative 4–10 point scales were used
to capture the employee’s perceptions on policies, processes, and
their impact on employee engagement.
Taking the marketing paradigm forward, companies did not
stop with gauging satisfaction. They found that satisfaction alone
was not enough to predict turnover levels. You could have a high
satisfaction and a high turnover.
Then additional dimensions were added. Employees were
questioned on the following:

• Intent to stay (loyalty).

• Intent to recommend the company to others (advocacy).

The measures for engagement started increasing. But what about

the business value?
One of the classic Harvard Business Review case-studies in
the 1990s related to the employee–customer–profit chain in the
American retail giant Sears. In the early 1990s, Sears had a new
CEO and were looking at large-scale organization transformation
(Rucci, Kirn, & Quinn, 1998). After a highly involved process,
the company expressed that it wants to be a “Compelling place
to work, shop and invest.” A team was entrusted with identifying
the total performance indicators for the same. After iterations,
they came out with a model connecting employee motivation to
business outcomes. The model was formed after rigorous statistical
analysis on a 70-question employee survey.
They identified “attitude about the job” and “attitude about the
company” as predictors of employee behavior. In turn, they found
that answers to six questions impacted “about the job”. These
covered,

• liking the work,

• pride in the company,
• impact of quantum of work on attitude about the job, and
so on.
100 Winning on HR Analytics

Similarly, response to four questions, such as

• confidence in the future of the company,

• understanding business strategy, and so on had an impact on
the attitude about the company.

That was the crucial first step in identifying the drivers

of engagement. The challenging piece remained on connecting
the engagement levels to business performance. Sears were able
to establish that a 5% improvement in employee attitude will
drive a 1.3% improvement in customer satisfaction resulting in
a 0.5% improvement in revenue growth. They were actually able
to predict the store performance just on the basis of engagement
survey outcomes. Subsequently, the model was revised to factor
in both business and employee measures.

The study that became a Harvard Business Review article was

the first to clearly identify a linkage between employee motivation
and corporate performance. The company estimated that it could
have added $200 million in additional revenues because of a 4%
increase in employee attitude.
Then War for Talent happened. Silicon Valley always had a
shortage of talented programers and in the dotcom era companies
outdid one another to offer benefits and perks to ensure longevity.
Even best employers started losing people. Employee turnover
became not just a long-term challenge, but one on which one
 Talent Engagement and Retention 101

needed instant improvement. This created its own set of questions.

What is the shortest (least-cost) path from being 2, 3 on employee
survey to 4 (on a scale of 5). Then RoI was needed for different
programs and their impact.
To an extent, the 2008 economic downturn took the sheen off
employee engagement. Large-scale layoffs became the new norm.
Sears happened in 1990s, before the “analytics” era was upon
us. But we have another equally powerful illustration from another
retail company, Lowe’s. Lowe’s had very clear linkages established
at the store level between sales promotions and improvement in
sales, linkage between revenues and shrinkage, etc. However, while
the executives felt there is a link between employee engagement
and business success, the models were not there.
The company set out to develop such a model using business
measures and the outcomes from their employee surveys. Often we
are curious to know about the following linkages:

• Can I increase revenues by increasing person days training?

• Can I increase retention by improving manager engagement?
• What are the three programs having the biggest impact on
employee engagement?

Lowe’s used a statistical modeling technique called SEM to vali-

date their hypothesis. Analysis showed that employee engagement
had a positive impact on customer satisfaction which, in turn, had
an impact on shrinkage numbers. They also discovered a two-way
linkage between manager engagement and employee engagement
(Coco, Jamison, & Black, 2011). This, in turn, helped them refine
a model that can help estimate the impact of engagement on busi-
ness. They found that higher engagement resulted in an average
4% higher volume per sale. Factoring in cost savings, they con-
servatively estimated a difference of a million dollar sales per
annum between the most engaged stores and the ones with least
engagement.
These are studies that clearly show a positive correlation between
employee engagement and company performance. Companies that
rely on human capital like Google go all the way to ensure the right
102 Winning on HR Analytics

talent is attracted and retained. However, cost constraints often

come in the way of long-term engagement of most companies.

Traditional Measures of Engagement

First and the biggest is measuring employee turnover. One could
argue that employee turnover is a part of workforce demographics.
While that is true, it is also the first measure everyone uses to
identify the strength of engagement.
SHRM defines employee turnover, as the “rate at which
employees enter and leave a company in a given fiscal year.”

Measuring Attrition
Unlike the headcount, attrition is always a percentage rate. It is
measured in annual, quarterly, and monthly intervals, though
sometimes companies in high turnover industries track manpower
leakage almost on a daily basis.
A simple formula for deriving attrition is:
Number of employees who quit during an year
Attrition rate = Average headcount for that year

When multiplied by 100, this gives the attrition percentage.

It is easy to find the number of employees who quit during
any given 12-month period. How do you arrive at the average
headcount?
It depends on the degree of accuracy you are looking for. A
basic formula would be:
 Talent Engagement and Retention 103

Headcount on 1st April + Headcount on 31st March

Average Headcount =
2

If you need greater resolution,

Average Headcount
(Headcount month 1) + (Headcount month 2) + f + (Headcount month 12)
=
12
Headcount of a month is usually calculated by averaging the
headcount on the first day of the month and the last day of the
month.

Month HC Exit

Apr. 1,200 15

May 1,250 18

Jun. 1,270 25

Jul. 1,450 45

Aug. 1,475 30

Sep. 1,550 20

Oct. 1,625 10

Nov. 1,660 10

Dec. 1,650 15

Jan. 1,700 18

Feb. 1,725 20

Mar. 1,725 15

What will be the attrition rate by the first method?

Attrition = ` j = 16.3%
241
1475

Now, if we average out the headcount across the year, then the
denominator becomes 1,525. The attrition rate then is:
104 Winning on HR Analytics

` 241
j = 15.8%
1,525

We are able to see a 0.5% difference in attrition on the basis of how

the averages are arrived at. While there is no right or wrong, any
average that considers more sampling is likely to be more accurate.

LTM or YTD?
There are no arcane terms. As far as attrition is concerned, there are
two different ways of arriving at an annualized number.
In most companies in India, the fiscal year (FY) runs from April
to March. On April 1st of the next year, you will be able to identify
the attrition rate for the previous year.
However, business is more dynamic to wait for a year to find
out the attrition rate. Often, updates are required on a quarterly or
monthly basis. Then what do you do?

LTM

LTM stands for last 12 months. In some places, TTM is also

used (trailing 12 months). This works on a rolling rate principle.
Suppose we want to know the attrition rate in August. We have
only seen 5 months in the fiscal. So, we use the attrition formula
for the preceding 12 months.

LTM attrition for August 2014

Number of quits
=
average headcount from September 13 to August 14

This is similar to how businesses calculate their revenue run

rate. While the company may have done $500 million last fiscal,
on the basis of their last 12 months’ revenue, they say our run rate
is 575 million dollar.
Advantage of going with the LTM basis for attrition calculation
is that you are always basing it on information available. There are
no assumptions being made here.
 Talent Engagement and Retention 105

However, suppose a company wants to set attrition goals for its

managers. The going rate is 18% and the company wants to bring
it down to 15%. So, it sets that all managers should keep attrition
at 14%. Appraisals have arrived. How to calibrate the performance
of managers?
One way could be to just consider the LTM attrition as of
appraisal time. This would include performance over the past 6
months. However, what if you want to only give importance to
performance in this fiscal?
Here year to date (YTD) comes into play.

YTD

One calculates the going attrition rate and extrapolates it for the
rest of the year.
Let us start with a simple example. Assume that the attrition
rate for a company at the end of April is 2%. Then using the YTD
method, we assume that the attrition for the rest of the year also
would be 2% per month. Total attrition on a YTD basis becomes
24%. (Not good for our managers!)
Let us take forward for a quarter.

Month Initial headcount Final headcount Attrition

April 800 820 12

May 820 845 10

June 846 866 9

What is the attrition for this quarter?

We see that the average headcount for April, May, and June is
810, 833, and 856, respectively. Average headcount for the quarter
then is 833. Company lost 31 employees.

Attrition rate for the quarter is ` j = 3.7%

31
833

On a YTD basis, we extrapolate it for four quarters by multiplying

by 4. The YTD attrition rate is 14.8%.
106 Winning on HR Analytics

This estimation happens to be a more conservative way of

forecasting attrition. In effect, the company is growing by 22
employees a month, while it is losing at the rate of 10.33 a month.
Extrapolating for the year, attrition could be

Annualized attrition = ` j = 13.3%

12 # 10.33
932

While rate extrapolation is a very quick way of forecasting, we

can project headcount growth and attrition and arrive at a more
accurate estimate as shown above.
YTD calculation is estimation. However, it is completely based
on current environment and does away with any impact from the
past or business actions from the previous year.
A usual mistake done by people is to take the rate for a period
and state that as the attrition rate. Just taking the rate for a period as
that for the year will set us up for shocks in the future. It is always
better to estimate the annual rate and then share it accordingly.
Both YTD and LTM have their backers and their uses. It makes
sense to be aware of both and use appropriately. YTD is a better
measure of short-term spikes, while LTM levels out such spikes
and presents a more realistic picture.

Employee Retention
Some companies are trying to move away from tracking attrition
and are focusing on their ability to retain employees. The primary
measure of retention is average tenure.

Average Tenure

Attrition finally is about employees leaving. What about those

who stay? This is where the length of tenure becomes important.
Objective of engagement is to get employees to stay with the
organization.
Let us take two companies. Assume that one has an average
tenure of 7 years, while the other has a tenure of 2 years. A
 Talent Engagement and Retention 107

company with a longer tenure has more stability and a better knit
culture. There could be three reasons why a company has a shorter
average tenure:

• The company is growing very fast and adding new people

rapidly.
• The company is new!
• There is a high amount of churn.

The last is what should be of concern. It would be useful to

benchmark across the industry and then see whether one’s own
company is able to retain people for a longer time.
It is also possible that tenures are longer in industries where
the demand for talent is low and value propositions are similar.
In general, government has longer tenures but that does not
automatically indicate highest levels of engagement. In a company
with longer tenures, change management could be difficult.

Lead Indicators of Engagement

While attrition is universally used as the measure for engagement,

it suffers from the following:

• It is at the fag end of the engagement process. It is too late to

do anything when an employee quits.
• An increase in company-level attrition happens sometime
after the engagement has dropped. It would be easier to do it
by measuring engagement on a real-time basis.

This is where surveys come in handy. Most organizations conduct

an employee survey at a periodic interval of 12–18 months. The
surveys usually focus on assessing the overall satisfaction level
and, with time, added elements of loyalty and advocacy.
The survey numbers were very helpful and independent of
which industry you are in—a satisfaction rate of 80% is always
more handy than 60%. When the scores drop from a survey to the
next, the leadership can also use it as a lead indicator of higher
turnover.
108 Winning on HR Analytics

A monolithic survey may make a lot of sense at an organization

level. However, in modern times when Moore’s law drives a lot
of things, 18 months really becomes a very long period of time.
Shorter, snappier polls are gaining currency. Even mood meters,
that track employee’s mood on a weekly basis, are seen as relevant.
By nature, surveys are confidential and one does not know about
who gave what feedback. This limits the predictability of engage-
ment. We may say that 30% of people in engineering are unhappy,
but we may still have to guess about who constitute the 30%.
Hence, there is need to create a cause and effect analysis for
lower engagement and for attrition.

Modeling for Surveys

It is beyond the scope of this book to write on how employee
surveys are designed and implemented. However, the following
points need to be kept in mind when using a survey for measuring
engagement.

1. Critical sample size: Often companies either have a small

sample or try to cover all employees. HR pushes employees
and managers to participate and shoot for 100% participation.
While a higher number looks good, it is not necessary for a
valid survey. It is more important to have the following:
A stratified random sample that ensures adequate
1. 
representation from all employee demographics. Let us
consider the following.

Level Actual number Survey number

1 20% 10%

2 30% 80%

3 50% 10%

 This survey participation, for example, has a very

different profile than what the company has. So, any
inference made on the basis of this data may not fit in
with what the company thinks.
 Talent Engagement and Retention 109

2. Statistically valid sample size: Companies try to have all

employees participate in the survey. That is good from
an employee involvement principle. However, accurate
assessment can be made with a much smaller sample
size. Using confidence interval, confidence level, and
stratification it is possible to arrive at reasonably accurate
modeling of very large populations using sample sizes in
the hundreds.
3. Input and output: Often dipstick surveys are used to gain
inputs on perception. This works as long as we do not
have one final outcome in our mind.
  Let us take an example of a restaurant. It is easy as
long as they ask for your feedback on a five-point scale
on the ambience, food quality, speed of service, etc. But
the moment they ask you about the overall satisfaction,
the dynamics change. What do you make out of the
following?
• Satisfaction with Ambience : 4.5
• Satisfaction with food quality : 3.8
• Satisfaction with speed of service : 4
• Overall satisfaction : 3.7

A manager sitting with this data will be confused. Some-

times, people just think taking an average of all scores
should result in the overall score. In such a case, the
average score here is 4.1. But the overall score is less
than the average. So what has happened?
Possibly, in framing the dimensions, a crucial element
that customers value has been left out. One can look at
factors like ambience, food quality, speed, etc., as inputs
that result in the overall satisfaction as an output. What if
value for money is a factor that is important but has been
left out? And the restaurant is perceived to be overpriced?
Does that impact the overall score?
Often, in employee surveys as well, one is faced with
this question. When formulating, have we identified all
drivers of engagement? Are there dimensions that are
overrepresented and underrepresented?
110 Winning on HR Analytics

This is where drawing up of a framework for engage-

ment becomes important. At a very high level, a company
can create a framework similar to the model shown below.

Lately, the models do not just stop with measuring

satisfaction but also result in conscious actions. Does a
satisfied employee:

• Express intent to stay for a longer time? (loyalty)

• Recommend the company to others? (advocacy)

Such modeling, of course, requires familiarity with

statistical tools and techniques. Statistics also takes us to
the realms of predictability.

It would be tempting to look at satisfaction, advocacy, and loyalty

as three axes that constitute a cube and analyze the patterns. It can
definitely be done, but with a caveat. The three measures are to
some extent interlinked and are not mutually exclusive.
 Talent Engagement and Retention 111

Lead Indicators and Predicting Employee Engagement

In manufacturing, two parameters of the shop floor are looked at
with great concern. One is tardiness/late coming and second is
absenteeism. Both, of course, have an impact on the productivity.
However, what connects them to engagement?
There is a theory based on withdrawal. It is said that people
tend to withdraw from spaces when they do not like being in that
space (Adler & Golan, 1981). Let us take it forward to the work
environment. In the flextime environment, one is never sure of
the starting time of the office. It would be a challenge to say an
employee is not engaged because he/she is not coming “on time” to
office. However, the following are observed to be lead indicators
of diminishing engagement:

• Change in work timings: An employee starts coming later

and later to work. 9:30 becomes 10, 10 becomes 10:30
without any increase in the hours at work.
• Increasing use of “work from home” options without any
personal need for the same.
• Unpredictability of attendance: The person may be taking
leave more frequently and often without any notice.

All the three indicate that the person is becoming more withdrawn
from the workplace and does not want to engage with the organi-
zation or colleagues. An Academy of Management Journal paper
from 2006 “How important are job attitudes? Meta-analytic com-
parisons of integrative behavioral outcomes and time sequences”
by Harrison, Newman, and Roth (2006) details out the correla-
tion between these factors as well as an outcome like employee
turnover.
On the other hand, we also tend to look at employees who are
supportive of their co-workers, talk well about the company, and
volunteer for discretionary work as those who are more engaged.
Organization citizenship behavior, as originally conceptualized by
D.W. Organ (1988), has the following key components:

• Altruism,
• Conscientiousness,
112 Winning on HR Analytics

• Sportsmanship,
• Courtesy,
• Civic virtue.

There are studies that also correlate organization citizenship

behavior with higher engagement as well as more positive business
outcomes.
So there is potential to leverage what is being anecdotally used
to track engagement to actually be used as a mainstream measure.

Employee Referrals as a Lead Indicator

As stated, employee surveys measure the “advocacy” factor as

well. While the advocacy scores measure intent to recommend
the company, does it happen in practice? Often large companies
say that more than 30% of their new hires come from employee
referrals.
Health of your employee referral program is a good indicator of
the level of engagement. Do you get a flow of recommendations
even when there are no positions? Or on the other hand, are you
hard selling your referral program and still not successful? Most
importantly, is the offer conversion rate higher when it is made
through referrals?
Each of this is a good lead indicator of the level of employee
engagement.

The Engagement Curve

Like economics, engagement also follows a pattern that has

been observed across companies. One typically sees a J-shaped
curve on engagement scores if we plot tenure on one axis and
engagement score on the other. Employees with 1–12 years of
experience show very high engagement levels and then it starts
dropping for next group of 10 to 20–22 years of experience, and
finally rises again for employees with 22–30 years of experience.
Why so?
 Talent Engagement and Retention 113

• Studies have shown that this happens because euphoria of

first employee after college and subsequent 2–3 job changes
in the next 10–12 years’ period keep the engagement levels
high.
• As the employee progresses on chronological age, his/her
cognitive faculties start developing very quickly and become
quite discerning about everything, including job and personal
achievements, and a sort of cynical outlook develops due to
more knowledge processing and exposure.
• After this stage, in late 40s and early 50s onwards, an
employees’ outlook towards life and work changes as the
employee becomes more receptive and appreciative of
whatever happens around him/her, presumably due to more
maturity dawning and because cynicism gets replaced by
security and likeness for everything due to short working
years ahead. This liking outlook shows in increased engage-
ment levels and hence that “handle” of the curve represents
the start or renewal of falling engagement levels once again!

Usage of Analytics to Fine-tune Action Items for Retention

Let us consider the technique of exit interviews. Companies

always had an exit interview format and it is still part of the
relieving formalities of an employee. Companies also analyze exit
interviews. However, exit interviews have ceded space in analytics
due to the following reasons:

1. Timing: An employee tends to be more transparent with his

emotions when he is still considering leaving the company.
By the time his last day comes, he is not interested in a
reversal of situations which could make him stay. Also, he
does not want to hurt anyone on the way out. So he reveals
only a portion of what really caused his exit. More often
than not, they say that their attrition is caused by “personal
reasons” effectively blocking off the company’s efforts to
identify the real reasons.
114 Winning on HR Analytics

2. Aggregation: The challenge for any reporting on engagement/

turnover is the heading given to causes. Summarizing results
in headers such as:
• Job-related,
• Personal reasons,
• Performance reasons.
Taking action at that level is often difficult.

The problems of aggregation dog employee survey outcomes too.

It is not unusual to have tables like the following as priorities for
action planning.

Level Junior Middle Senior

Priority Career growth Work-life Collaboration

Compensation balance Communications
Training Job rotation
Benefits

There are only a few things that can be done to take action
on a sustained basis. Also, this does not differentiate in terms of
employee expectations at different businesses, locations, etc. Here
is where supplementary analysis comes in handy to do the fine
tuning. Let us consider the following situation:

BU A B C D E F

Attrition 14% 9% 10% 20% 12% 16%

Let us suppose that the company attrition rate is 12%. So, the
business units (BUs) A, D, and F are higher than the company.
A lot of literature exists connecting the impact of manager with
employee engagement. So can we clearly say that the managers of
BUs A, D, and F should be relieved of their managerial roles and
fresh blood be brought above them? Or there are other factors we
need to look for?
Let us consider a few factors in the spirit of our deep and wide
analysis (assume this is a technology company):
 Talent Engagement and Retention 115

• Manager maturity (as indicated by the number of years in the

role),
• Demand for skillset,
• Team average tenure with company,
• Employee survey scores.

To simplify it, let us segregate the data.

Factor A B C

Manager maturity

Skillset demand

Team tenure

ESAT score

Notes: Manager Maturity ranges from high to low. Black for high, white for average and
gray for low.
      Demand for skillset: Black for less demand, white for moderate demand and gray
for high demand.
      Team Tenure: Black for high tenure, white for moderate and gray for low.
      ESAT Score : Black for high, white for average and gray for low.
      These details make it more complex and do not point fingers on managers at all!

Business Unit A
BU A has a manager who has been playing that role for a while. The
demand for team’s skillset is moderate. The team has been with the
company for some time and even the employee satisfaction scores
are reasonable. There would be a need to examine the manager’s
personal style and make corrective actions.

Business Unit B

The demand for skillset is moderate and the ESAT scores are
reasonable. However, the team itself has not worked in the
company for a good length of time. They may be still settling in.
Also, the manager has just moved into people management. A
116 Winning on HR Analytics

right approach would be to coach and develop the manager so that

he/she becomes better.

Business Unit C

While the manager has reasonable experience, he has also delivered

high employee engagement scores in the survey. The attrition
should not be high, looking at it conventionally.
But let us consider the external environment. The demand for
the skillset of the team is very high. This, in turn, leads to far higher
attrition than the company is enduring otherwise. High turnover, in
turn, creates a churn resulting in low tenure of employees. A right
approach would then be to re-examine policies to make sure the
unique demands of the BUs are met.

Factor A B C

Manager maturity

Skillset demand

Team tenure

ESAT score

Action Check style Coach manager Change policies

Note: Manager Maturity ranges from high to low. Black for high, white for average and
gray for low.

Three units, three different challenges—only by using analytics

will HR be able to take decisive action. Using templates for
solutions will result only in frustration.

Building up for Retention Predictability

It is fascinating that two people from the same background join a

company and one leaves within the year, while the other goes on
to become CEO after 25 years! A company cannot be so different
for two people?
 Talent Engagement and Retention 117

A lot of literature exists on what helps higher employee

engagement, leading to retention. Gallup’s Q12 based on their
research has become iconic in this field, even though with time
some elements like “Have a best friend at work” or “I received
recognition in the last 10 days” have become open to criticism.
(Crush, 2009)
Let us visualize an empirical model to envisage what all go into
building an algorithm for predicting retention.
Let us assume that two people join with exactly the same
background. What would be the factors impacting the initial
engagement?

• Have I received the right fit in terms of level and

compensation?
• Are people nice to interact with?

Within a month, the impressions will start getting formed on the

following:

• Is my manager good to work with?

• Does the company have a legitimate strategy?
• Do colleagues talk well of the company?
• Are my job expectations clear?

After 6 months, the following impressions start getting formed:

• Is the culture democratic? Can I express myself without

fear?
• Is my manager enabling me to be successful? Is he/she
approachable?
• Am I feeling stretched?
• From a pay and benefit perspective, is the company
delivering what it had promised, especially bonus and
variable payment?
• How easy or difficult it is, to get things done here.
• Is my team composed of competent staff?
118 Winning on HR Analytics

After 1 year, the following takes precedence:

• Have I started achieving what I was hired for?

• Am I getting visibility to senior management?
• Am I getting selected for task forces?
• If in a remote center, do I have the freedom to visit HQ when
needed?
• How much discretion do I have to spend my budgets?
• How much discretion do I have in hiring my team?
• Is there any investment being made into my development?
• Did I get a fair performance review?
• Did I get a fair pay increase?
• Is this an exciting area to work?

After 2 years, a person might start having these questions:

• What is my visibility into the next higher position?

• How strong are my networks within the company? Do I like
working here?
• Does it ring a bell, when I say I work here? Is there social
recognition?
• Am I able to get time for family when required?
• Does senior management rely on me to deliver results?
• What are my long-term wealth generation prospects?
• Am I being recognized adequately for my efforts?

This is one shot at identifying the factors that are important for a
person in managerial levels/mid-career. The factors could be very
different for someone in the early stage of their career. These could
vary slightly for someone at senior level of leadership. (It might
just come down to working relationship with CEOJ)
For the sake of simplification, let us summarize as in the
following table:
This is just one part of the modeling. There are other questions
of a more personal nature like:

• Proximity to native,
• Quality of life in place of work,
 Talent Engagement and Retention 119

Level Important factors Good to have factors

Junior Training Work culture

Cutting edge work Supervisor quality
Pay
Benefits

Middle Career growth Work-life balance

Responsibility Benefits
Work culture Cutting edge work
Management quality Higher education
Pay

Senior Success on the job Perks

Peer respect
Leadership vision
Influence ability
Wealth generation

• Job and locational preference of spouse,

• Nature of financial commitments.

All these of course fall into the mythical bucket of “personal

reasons” as stated in exit interviews!
HR organizations model retention with the sincere belief that it
is all under their control. It is not. External market reasons play as
much a role in retaining an employee, especially at junior levels,
as intrinsic reasons.
We have organization factors, individual factors, and market
factors. For simplicity’s sake, let us give a weightage for each
factor, so that these add up to 100. Let us also assess an employee’s
current level of engagement using a 1–3 scale, where 1 is low.

Work Pay and

Factor Learning quality benefits Manager Team Personal

Weight 15% 25% 20% 15% 10% 15%

EE 1 2 3 3 2 3 2

EE 2 3 2 1 1 2 3
120 Winning on HR Analytics

Let us multiply to arrive at strength of retention:

EE 1 = (0.15 × 2) + (0.25 × 3) + (0.2 × 2) + (0.15 × 2) + (0.1 × 3) + (0.15 × 2) = 2.55

EE 2 = 1.95.

The range for scores is between 1 and 3. One might classify as the
following segments:

• 1.7: High risk,

• 1.7 to 2.4: Moderate risk,
• 2.41 to 3: Low risk.

Based on intrinsic reasons, EE 1 is low risk and EE 2 is moderate

risk.
This has to be aligned to the market condition. We might again
use a 1–3 index, with 1 as high demand and 3 as low demand. In
which case, we take the score again and multiply. Assume that
EE 1 is in a low demand area while EE 2 is in a high demand area.
Then, 9 becomes the desired score (3 on engagement and 3 on
demand). Treating intrinsic and extrinsic factors equally, we arrive
at final scores. EE 1 = 2.55 × 1 = 2.55, whereas EE 2 is 1.95 × 1 =
1.95.
Even when the intrinsic motivation is high, the risk of EE 1
leaving is higher, as the demand is much higher in her area. On the
other hand, EE 2 is at a lower level of engagement but not a flight
risk, as in his work area, demand is less.
At an enterprise level, such fine-tuned modeling is difficult.
However, data science and big data come in handy here and help
us model on far more parameters that are important. For example:

• Tata Hotels recruits primarily from tier 2 cities in India as

they find their attitude to be better.
• TVS group looks at children of its employees from group
companies favorably for entry-level hiring.

On the one hand, we use data scientists to create the algorithms

that create a right fit equation for predicting turnover. On the
other hand, we also need to capture the data that is crucial. We do
 Talent Engagement and Retention 121

have information from workforce data. We can create the external

demand factor based on actual data. Employee’s own perception
and needs, have to be arrived at on the basis of the following:

• Surveys that track individual preferences,

• Strength of connections, participation, etc.

Predictive Modeling for Attrition Analysis

As technology and analytic tools have advanced, organizations are
employing sophisticated statistical models to mine attrition data
and understand at the granular level what trends and patters emerge
with what probability of success. Many companies have employee
data like performance ratings, promotion history, training programs
attended, engagement data, salary data, tenure on various positions/
roles, age, leave, etc., and added to this, data from social media can
be used effectively to find which employee will quit. Also data like
traits of employees who have been successful performers in the
company can be used to profile employees who will not fit with
these traits and, hence, pose the risk of quitting company. Here are
some of the popular predictive models used for attrition analysis:

1. Descriptive statistical techniques: HR function has lot of

historical data and an easy way to use this data for predic-
tive analytics is by using descriptive statistics approach.
Descriptive statistics methods like forecasting with past
data (stationary series), moving averages and exponential
smoothing can be used at the very simpler level to predict
attrition for the next month or quarter at aggregate level.
Even seasonality in attrition—which is a common phenom-
enon in some organizations—can be factored by exponential
smoothing forecasting method to predict attrition in the next
month or quarter due to seasonality factors.
2. Regression analysis: To further improve predictive validity,
a commonly used statistical tool for predictive analytics is
regression analysis. Regression modeling frequently used
includes:
122 Winning on HR Analytics

a) Ordinary least square (OLS) regression: This is used

to study the relationship between independent and
dependent variable using equation y = mx + b. For
example, regression of attrition and engagement will
give result like – A (Attrition) = –2E (Engagement) +
2,000, which means that for every 2 points increase in
the engagement score, attrition will go down by 2 points
– 2,000 – 2 = 1,998. Also the graph plot is a linear line
indicating variables fit on the line and nothing more.
However, this model does not tell whether change in
engagement levels will lead to the employee quitting
or staying. For this another type of regression modeling
will work.
b) Logistic regression: This method tells very clearly
whether employee will quit or not—like 0 or 1
modeling—assuming 0 for stay and 1 for quit. Here the
graph plot will be like “S” curve which highlights when
attrition will go up or fall.
3. Artificial neural networks: These are predictive models
based on brain analogy and work on the basis of experience
or training as in the case of brain. Neural networks are ideal
for working with complicated and imprecise data to identify
patterns and trends.
4. Classification methods: These include decision trees and
memory-based reasoning/case-based reasoning. These are
like algorithms which are suitable for identifying patterns in
non-numerical data.
5. Clustering: It is another model to predict groups which
are similar or closer to each other based on some traits or
attributes. This tool can be useful in identifying clusters of
employees who display high or low relationship between
job attitudes and turnover. Clustering will show patterns
like dense areas (high correlation) and spare areas (low
correlation) based on attributes.
6. Survival analysis: This technique is borrowed from medical
sciences where it is used to predict survival of a disease-
affected patient based on medicine used for the treatment.
 Talent Engagement and Retention 123

Survival analysis is an analysis of time to predict when an

event will happen and the time elapsed before the event
happens. In employee attrition, it helps in pointing out
when the event (attrition) will happen with what probability
and how long will it take for the event to occur with what
probability. For example, analysis may show that there is
80% probability that 10 employees will quit after 8 years or
50% probability that 6 employees will quit after 5 years, and
so on. Survival analysis includes analyzing the following
and taking action:
a) When an employee will quit, at different time period in
months or years
b) What is the probability of quitting taking place for that
time period?
c) Take action before the predicted event happens.
 8
Measuring and Managing Competencies

C ompetencies are the connecting blocks of mature HR prac-

tices. People capability maturity model designed by the
Software Engineering Institute provides for five levels of matu-
rity in HR practices. Practices such as recruitment, training, and
compensation occur at Level 2.
 Measuring and Managing Competencies 125

As the schematic shows, business and HR are linked using

competencies as defined in Level 3. The business expectations are
converted into competencies needed from the workforce. Then
the competencies are embedded into workforce planning, career
development as well as into recruitment, training, promotions, etc.
At a higher level of maturity, the processes are made predictable
by baselining process performance and then predicting based on
the trends.
Emerging practices in HR rely on competency-based assess-
ments to not only arrive at the best individual decisions but also:

• Decide on the focus areas for L&D,

• Hire effectively based on competencies that drive retention
and performance.

Competency Baselining
Most companies have identified a set of competencies that are
mapped onto the roles. A table could look like this.
ROLE: Testing Lead

Competency Level

Communications 2

Collaboration 2

Analytical ability 1

Testing 3

Next step is an assessment of all role-holders. This could be

done using a self and manager feedback, 360° survey, or a proper
formal assessment. In general, at the end of the assessment, an
employee gets a score. Expanding the above, the assessment could
look like:
126 Winning on HR Analytics

Competency Level Score

Communications 2 3/5

Collaboration 2 4/5

Analytical ability 1 3/5

Testing 3 3/5

The company can then consolidate the score from all such
assessments and arrive at a baseline score for competencies as
follows.

This can then be used to share with the individual employee

where he/she stands with respect to the mean score of all
role-holders.
Such a comparison will still need to be positioned appropriately,
as the averages can be in fractions and the employee can only score
in discreet numbers. As it is done elsewhere, percentile scores can
also be used to share the findings. This is as far as the individual
is concerned.
The organization can then perform a regression of the
competency data with performance data. How is the competency
profile of the top 10% performers different from that of the others?
 Measuring and Managing Competencies 127

Are the differences across all competencies or only one or two

competencies account for the performance variation?
Research shows that a different level of proficiency in just three
competencies out of a larger basket can influence performance by
more than 75%. Analytics helps us to find what these competencies
are. Using the simplified table above, assume we get the following
data.

Competency Average Top 10%

Communications 3.25 4.3

Collaboration 4 4.1

Analytical ability 3.75 3.9

Testing 3.6 4.3

This data clearly shows that on collaboration and on analytical

ability, the scores do not differ much. Top performers are assessed
at a higher level on communications and testing. It is possible that
proficiency in these has a greater impact on performance.
The company needs to validate this after considering other
variables such as nature of work and quality of supervision, etc.
Let us assume that even after that, testing and communications
emerge as differentiators.
128 Winning on HR Analytics

Usage of Competency Baselines

The first usage of such baselines happens in training. The company
can focus its training efforts on improving the communication
skills of all role-holders. It can also provide opportunities for
all test leads to get better in their functional skills by providing
development opportunities.
Secondly, the employees can not only drive their individual
development plans but also assess them against the requirements
of the higher role. While it has not become a mainstream practice,
a lot of companies do perform potential appraisal. Let us say that
the employee is looking for his/her next promotion to that of a
manager. Let us do a competency comparison.

Lead Manager Threshold score

Competency level level Score for the next level

Communications 2 3 3/5 4

Collaboration 2 2 4/5 2

Analytical ability 1 2 3/5 4

Testing 3 3 3/5 3

The employee needs to improve proficiency in communications

and analytical ability to be considered for a promotion. The last
column says about the threshold level assessment on the lead
level that can be extrapolated to the entry level proficiency for a
manager. Accordingly, we get a gap of 1/5 on both competencies.
This essentially means that the employee needs to display the
behaviors expected from these competencies more frequently to
be considered for a move up.
This is a simple way of doing it. For critical roles such as of a
manager and director companies often do a stand-alone competency
assessment to decide on promotions.
Thirdly, competencies can be tracked at a team level as well, by
aggregating the competency assessment scores. Creating simplistic
data, we get the following.
 Measuring and Managing Competencies 129

Team 1 Team 1 Team 2 Team 2

Competency lead manager lead manager

Communications 3.8 4 3.4 3.9

Collaboration 4.1 4 3.7 4

Analytical ability 3.2 3.4 3 3.7

Testing 3.6 3.8 3.6 4

This information can be correlated with project/team

performance. Let us assume that the team is tracked on:

• Percentage of on-time delivery,

• Number of escalations,
• Attrition percentage,
• Defect rate.

Figure 8.1 is an illustration of the same. The graph maps a

measure called “cost of quality” used in measuring quality of
programming. The measure has been mapped on one axis. On the
other axis, we have two indicators:

• Tech index: This is based on the cumulative competency

scores such as programming, analysis, and testing.
• Competency index: This is a composite of competency
scores across technology and behavior.

The chart shows cost of quality scores across different projects

against their tech and behavioral competency indices. The progres-
sion clearly shows that higher the competency scores, lesser the
cost of maintaining high quality.
It is possible to compare the competency scores with the business
measures using correlation and regression to arrive at the following:

• Whether competencies are impacting team performance.

• What is the impact?
• What needs to be done to enhance the competency level of
the team?
130 Winning on HR Analytics

Figure 8.1 Correlation: COQ to Competency Baseline

Source: Created by Samit Deb.

We are using competencies to assess and develop employees.

However that is just with existing employees. What about new joins?
Should we not also assess them against desired competencies?
Mature companies use competency-based hiring instruments.
Assume that communication skills is the biggest predictor of
on-the-job performance. Then, it becomes important to spend
some valuable time in the selection process assessing whether
the candidate displays proficiency levels commensurate with top
performers of the organization. This helps in ensuring that the right
profile candidates are selected and they start performing earlier.

Leadership Development
Competencies form the base of all leadership development
activities. Assessment centers are performed using leadership
competencies and development plans arrived at.
How do leaders of one company compare against best in class
leaders? This information is an added input when a company is
trying to be competitive with the market leader. Assessment
 Measuring and Managing Competencies 131

companies have the global trends on the average scores as well

as on the scores for top performing companies. A radar chart then
helps baseline the existing competency levels against industry-
leading competencies and then use that to fine-tune development
plans.
A radar chart, as follows, helps in this benchmarking.

Using Competencies in Talent Acquisition

As stated earlier, a company can move to a higher level of
HR process maturity by integrating competencies to all HR
policies. While it is one thing to assess and map competencies
to performance, it would greatly benefit to use the competency
information for hiring new talent into the organization.
Often companies have 12–15 competencies identified for
each role. Even if we prioritize a few for each role, it would
still mean 5–6. Assessment centers are a validated method for
objectively assessing potential employees. However, it is still an
implementation challenge to use assessment centers for all roles,
especially those that are done in high volume.
Typically, for entry-level roles, interviews are used for selection
and these do not exceed more than an hour or so. In which case,
how does one prioritize what competencies to focus on?
This is where correlations help. Employees are assessed
against different competencies. As shown in the example above,
132 Winning on HR Analytics

Average score for all Average score for

Competency role holders (1–5) high performers

Communications 3.7 4.5

Team working 4.1 4.2

Innovativeness 3.1 3.8

Analytical ability 4 4.2

Client orientation 3.5 3.7

the competency profiles for high performers are called out. The
crucial 2–3 competencies that differentiate high performers from
the rest are isolated. The limited time in interviews can be focused
on assessing these competencies to arrive at better quality of hires.
As the table above shows, the biggest differences between
average performers and high performers stem from the compe-
tencies of innovative thinking and communications. When one is
interviewing, it would aid to focus on these two competencies.
 9
Optimizing Compensation and Benefits
for High Performance

C ompensation and employee welfare costs are deemed

important enough for it to deserve a line for itself in the annual
P&L report of companies. These tend to be one of the top five costs
for any organization. Let us consider the following companies.

Employee cost/
Company Segment Headcount Revenues

Infosys IT services 165,000 57.50%

TCS IT services 300,000 55.50%

Marico FMCG 1,938 6%

Tata Motors Manufacturing 65,500 9.2%

One finds that there is a wide difference in employee costs

across the three segments represented here. Employee costs
account for more than half the costs in IT services. On the other
hand, in Tata Motors it is only 9% and in Marico, it is even less.
In a manufacturing company like Tata Motors, oftenthe cost of
raw materials is the highest amount costs; in services companies,
billing hours of people is the primary raw material. In fast-moving
consumer goods (FMCG), cost of materials and cost of sales and
marketing outstrip employee costs.
134 Winning on HR Analytics

Let us now detail out the table a little more.

Employee Employee
cost/ cost/Net Employee
Company Segment Headcount Revenues profit cost/FTE

Infosys IT services 165,000 57.50% 2.7 1,800,000

TCS IT services 300,000 55.50% 1.8 1,514,000

Marico FMCG 1,938 6% 1.7 1,469,040

Tata Motors Manufacturing 65,500 9.2 1.5 3,200,000

Source: Derived and simplified from annual reports 2014–2015.

In all four companies, the employee costs outstrip the net profits.
Even when the cost is as low as it is in Marico, it is still higher
than the profits. So, any path to higher profitability necessarily
incorporates an optimization of all employee expenses.
The last column is interesting because it is on the higher side, for
an average Indian employee. All this indicates is that Infosys, TCS,
and Tata Motors incur a significant payroll and benefit cost over-
seas, that when averaged and aggregated looks quite high. Marico
possibly pays more given its lesser headcount. In manufacturing
and FMCG, there is also contract labor and how they are expensed
is something that is outside the scope of this book.

Business Levers of Compensation and Benefits

Often in areas such as training and employee engagement, we start
off from an activity measure, derive the cost, and then try to assess
the business impact. On the other hand, compensation is an area,
where the costs and business impact are very well articulated.
As mentioned above, the cost of compensation and its impact
on the overall profitability is the first business lever. Not long
back, General Motors, the venerable American auto manufacturer,
went into bankruptcy because of an employee benefit; in the olden
days, they had guaranteed lifetime healthcare benefit for all. Over
 Optimizing Compensation and Benefits for High Performance 135

a period of 50 years, the cost of administering this benefit alone

brought the company to its knees.
Compensation is an area that is also governed by government
regulations, which set the minimum wages. Compensation strategy
becomes a fine art of balancing between the regulations, market
pay as well as employee performance.
Incentive compensation is used as the default tool to align
the sales goals with sales compensation. A lighter shade is the
investment on rewards and recognition programs. Stock options
and long-term incentive programs are used to align the long-term
growth of the company to employee motivation.
So, while employee expenses are an important line item in
financial statements, there is an equally big expectation to achieve
business goals using the motivation of incentive schemes and
reward programs. In effect, if HR has to stand its ground from
the CFO just looking at the people investment as expenses, an
analytical approach is de rigueur.

Organization Structure and Cost of Management

Management is important. At the same time, it is expensive. And
as the levels increase, the compensation rises exponentially. Let us
look at this example.

Role Number of people

Engineer 48

Senior engineer 27

Tech. lead 13

Project manager 6

Manager 3

Sr. manager 2

Director 1

Total 100
136 Winning on HR Analytics

Span of control will be the first measure used to identify whether

this is an effective structure or ineffective. Let us assume that
levels above project manager are managerial roles. Here, we have
6 + 3 + 2 + 1 headcount in such roles; 11 managers for 100 employ-
ees give us a manager span of control of 9. It is a reasonably good
number.
Now, let us include the average compensation costs for these
roles as well.

Number of Average salary Total

Role people pa compensation

Engineer 48 400,000 19,200,000

Senior engineer 27 700,000 18,900,000

Tech. lead 13 900,000 11,700,000

Project manager 6 1,200,000 7,200,000

Manager 3 1,800,000 5,400,000

Sr. manager 2 2,200,000 4,400,000

Director 1 3,000,000 3,000,000

Total 100 698,000 69,800,000

Let us relook at the ratio now. The average salary of an

employee is 698,000. The average compensation of an individual
contributor is 565,909. On the other hand, the average salary of a
managerial employee is 16, 66,667. In other words, every manager
costs nearly three times an individual contributor and, hence, direct
management costs would account for 33% of the payroll. With due
respects, managers administer value; they don’t create. In an IT
services company, all value comes from individual contributors
like software developers, architects, and sales people.
This is just for 100 people. Any mid-size organization employs
people in thousands. To that extent, the cost of management
increases proportionately. And we are not even talking about the
leadership layer.
 Optimizing Compensation and Benefits for High Performance 137

Percentage of total
Level Average Salary pa payroll

Individual contributor 565,909 71%

Management 1,666,667 29%

This graph clearly shows the difference. Engineers maybe 48%

of the headcount, but they cost only 27% of the payroll. On the
other hand, the director and senior manager level is just 3% of the
headcount, but 12% of the payroll.
Let us complicate things a little more. Let us improve the
headcount by 10 times. There are not only going to be more
senior managers and directors, but even people at the level of
vice president. For argument’s sake, let us say a company needs
1 vice president, 3 senior directors for every 1,000 employees,
while the structure for every 100 employees is replicated as above;
1 director for every 100 employees. You need to pay vice presidents
really well!
138 Winning on HR Analytics

Average Total
Role Headcount salary pa compensation

Engineer 480 400,000 192,000,000

Senior engineer 270 700,000 189,000,000

Tech. lead 130 900,000 117,000,000

Project manager 59 1,200,000 72,000,000

Manager 29 1,800,000 52,200,000

Sr. manager 19 2,200,000 41,800,000

Director 9 3,000,000 27,000,000

Sr. Director 3 4,200,000 12,600,000

Vice president 1 6,000,000 6,000,000

Total 1,000 709,000 709,600,000

Percentage of total
Level Average salary pa payroll

Individual contributor 565,909 70%

Management 1,666,000 28%

Organization leadership 4,650,000 2%

The drop in the proportion of individual contributor payroll

has been taken up by the leadership layer. For the cost of a vice
president, we can employ 15 engineers and they generate revenue!
This is simplistic. Vice presidents not only hold the fort but also
strategize for the company. They not only are busy in meetings
with each other but also with customers, which help the company
get more business that keeps engineers occupied—fair point.
However, the takeaway from this example is that:

• Span of control is an important measure.

• Dividing the payroll costs by management levels provides
greater insight.
 Optimizing Compensation and Benefits for High Performance 139

• Performance measures need to be tighter before increasing

the salaries of leaders.

Google famously once tried to do away with all managers and tried
to create a truly collegial workplace. However, they dropped the
experiment once the engineers started coming to the founders for
things like travel reimbursement. Managers were reinstated. There
are several experiments in organization structuring and design.
Maybe one day companies like Morningstar (a tomato processing
company), who have been successful without a single manager,
could become the norm. For now and the near future, management
remains integral to organizations.

Traditional Measures of Compensation

In general, two measures take preponderance in public discourse.
First is the compensation quartile positioning of the company, when
compared to a peer group. Second is the annual average percentage
increase in the base pay. What is also tracked by employees is the
percentage payout of incentive/variable payout. The latter two are
more popular than arguably any other measure in HR.
Mature organizations also track the following:

• Cost of defined benefit plans like hospitalization insurance.

This is usually the most expensive benefit and undergoes an
elaborate annual renewal process.
• Coverage of reward and recognition programs. How many
unique employees were covered at least under one of the
programs?
• Compensation on a cost-to-company (CTC) basis. Employees
do not really consider the investment into benefits like
provident fund (PF) as something that is contributed by the
company. In reality, at least 15% of the cash compensation
is contributed through benefits. Unless this is tracked at
the organization and employee level, there is no proper
appreciation of this investment.
140 Winning on HR Analytics

Organizations try to maintain or reduce the total cost of

compensation year on year. It is expected that any investment into
salary reviews or introduction of other benefits should lead to a
corresponding decline in employee attrition. Attrition of employees
based on their compensation quartile as well as percentage increase
in salary is tracked.
In addition to this, the traditional measure of comparatio is also
adapted by companies. Comparatio is used to assess the positioning
of the employee’s compensation with respect to the market pay
for that position. Market benchmarking allows us to establish the
minimum, median, and maximum for each position. Comparatio
then is a simple calculation as follows:

Compensation of the employee

Market mid-point for the position

A comparatio of 1 means that the employee is being paid the

market rate for the position.
It is possible to arrive at the group comparatio for different roles
and assess the market positioning of the role.

How Far Does Annual Compensation Increase Help?

No questions asked. Organizations do a benchmarking of their
competitors every year and based on business needs decide
on the comparatio to be used. Then, the percent adjustment
needed to achieve market positioning is arrived at. Employees
are then positioned across the median, based on their individual
performance appraisal ratings.
This usually is the annual merit increase process. However, it
so happens that in several organizations, the attrition peaks usually
after the salary increases. Some of it has to do with the employee
having a certain expectation and leaving when it is not met.
However, it is not unusual to find employee grievances around
fairness and internal equity. Neither is it unusual to see consistently
good performers not necessarily topping the pay charts.
 Optimizing Compensation and Benefits for High Performance 141

Very often one comes across employee angst around “I

joined on the same day as XYZ, my performance ratings have
been superior to him and yet, how is he making more money
than me?”
Such oversight happens, because companies lack an analytical
framework to visualize what the compensation distribution should
be and what it is. An annual process helps us make increments but
the starting point of it should be internal inefficiencies in salary
distribution.
Differences exist between the compensation of otherwise
equivalent employees due to the following reasons:

• Different salary at the time of joining.

• Variations due to salary increases in different years. One
person might have been rated “Outstanding” but received
little or no raise as that was a bad year for the economy. On
the other hand, his peer could have been rated “Outstanding”
in the following year, when the economy rebounded and the
company gave out good substantial raises.
• The employees were under different managers. Each had a
different peer group for comparison, leading to one manager
giving an increase on the higher side of the range.

Such factors add up over a period of time and result in variations

that are large enough to make the employee feel less engaged.
How does one address this using compensation analytics?
The solution lies in moving away from looking at employee
compensation as a one off to reviewing it holistically on the basis
of the following:

• Role spans and an employee’s positioning on it based on

tenure and performance.
• Employee’s compensation progression based on performance
in the span.

Let us say that a company has a role with a role span of 3 years.
Employees in this role will attain full proficiency in 3 years, when
142 Winning on HR Analytics

maintaining “meet expectations” standards. If the employee has

been consistently exceeding expectations, he/she would attain full
proficiency in 2 years. Let the company have four performance
ratings: exceptional (E), exceeds expectations (EE), meets
expectations (ME), and does not meet expectations (DNME).
For us to review this against employee compensation, we need
to map the cumulative performance of the employee in the role
against the current compensation. We are looking at performance
in band as well as compensation as a continuum and not stand-
alone events.
To start with a simple assumption, let us just examine the
scenario for employees who have been consistently rated E, EE,
ME, and DNME. Then the actual plot would be something like:

This shows that the person who is performing consistently lower

than expectations, (if he/she is lucky enough to retain the job) is
definitely at the lower end of the pay scale. For all the other three,
there is a clustering at the top. After 2 years, the consistency of
performance does not seem to be a big differentiator. This is slightly
exaggerated, but in the long run there does exist a clustering of
compensation.
 Optimizing Compensation and Benefits for High Performance 143

How should this graph look like? An average performer should

be at the mid-point of the role when their tenure equals the mid-
point of the salary range for that role. The exceptional performer
should be at the top end, in Q4 by the 3rd year.
The graph then should resemble this:

As is visible, the model also makes sure that the employee

compensation is not solely determined by performance on the job.
There could be two people rated as EE; one with 2 years’ experience
and another with 6 years’ experience. It makes little sense to give
both of them same increases, even though the performance rating
has been the same.
This is where the role span comes into play. Compensation
increases linearly till the 3-year mark. After that, a person
continuing on the same role could be delivering diminishing
returns year after year. As such, the pay increases should be muted
with time.
Is it possible to create a template that helps in this? Such a
template should be able to include cumulative performance levels
as well as tenure in the role. An illustrative table follows.
144 Winning on HR Analytics

The cumulative performance level is taken as the marker. Here,

we are looking only at four combinations that serve as guidelines.
An employee could be ME, E, and EE in 3 years while another
could be EE, EE, and ME. These scores fall within the overall
range and can be calibrated. There are four annualized performance
ratings, resulting in 24 combinations.
Such a framework based on existing compensation information
is useful in validating the compensation strategy. This exercise
should be conducted when the compensation planning is initiated.
Using this matrix, the compensation manager can identify outliers
as they exist across the company.
At present, most companies do not consider the cumulative
performance evolution. Then outliers are handled as exceptions and
actions are taken based on managerial discretion. A systematic top-
down approach using tenure in role and cumulative performance
rating allows outliers to be highlighted early and actions to be
taken in a timely fashion.
Often, one is at a loss to understand why employees leave,
even after getting a good raise. Only after an investigation is HR
able to identify systemic issues that have created a perception of
unfairness. Articulating fairness as a compensation goal and using
tools to ensure fairness in action reduces post-salary increase blues!

We Are a High Performance Organization.

Are You Sure?
While purists may scoff at the format as well as the “auctions”
used to hire players, the Indian Premier League cricket tournament
 Optimizing Compensation and Benefits for High Performance 145

promotes merit. It rewards the pace bowler, big hitting batsman,

and the nifty all-rounder based on their performance. Teams bid
for players and the annual pay packets can be in millions of dollars.
A team would have two opening batsmen; one costing 10 crores
per annum and the other, 50 lakhs; A 20X variation. Players do
not mind it much as they feel that with consistent performances
they will be able to command premiums as well. The IPL does
not classify players into opening batsman, opening bowler, and
spinner and then arrive at a median salary.
Organizations are somewhat different from sports or enter-
tainment. It is said that the best programer can be 30 times more
productive than the average programer. However, an organiza-
tion finds it difficult to create a talent market due to the following
reasons:

• The deliverables are interlinked and it is not easy to

completely isolate one person’s contribution.
• The employees have a 30-year career, when compared to a
star athlete, who has a 10–15-year career.
• The expectation is for a few stars but many solid performers.
The compensation base is set on an average employee
delivering average performance.

However, companies rigorously rank their employees on a

performance curve with the intent of allocating salary increases on
the basis of performance reviews. This is showcased as one of the
primary vehicles of meritocracy.
Even allowing for all variables, just how much more should
be the incentive for the top performer? Is there a top performer
band?
At the end of normalization, most employees are in the middle
of the curve. After benchmarking, the average salary increase is
arrived at, so that the median is fixed first. If there is a 10% gap
between market salary and internal median, then it is decided that
there would be a 10% adjustment. What about the top performer?
Does he/she get 5% more, 10% more?
146 Winning on HR Analytics

In practice, managers and organizations are very comfortable

with central tendencies. Once the average increase is arrived at,
a premium of 5% or so is added to the salary increase for the top
performer. A compensation matrix looks like this:

Performance rating E EE ME DNME

Percentage workforce 10 40 45 5

Salary increase 16 12 8 0

The increases are in steps of 4%. An employee, who has

outperformed everyone, ends up getting just 8% more than the
average performer. In informal discussions, employees say that
the effort required to achieve the top rating is just not worth
the returns. Drawing a comparison chart, one gets the attached
picture:

Please note that the ideal distribution here aims to create

increasing levels of differentiation between each performance
level. This can be achieved by focusing not only on the median
number but also on the difference between each performance
level.
 Optimizing Compensation and Benefits for High Performance 147

Performance rating E EE ME DNME

Percentage workforce 10 40 45 5

Salary increase 16 12 8 0
percentage

Desired salary increase 22 14 8 0

percentage

Starting point for all this will be the simple scatter plot of
employee compensation against performance.

Valuing Benefits Using the CTC Statement

Consider two friends discussing their salary.
Friend 1: How much do you make?
Friend 2: I make 12 lakhs.
Friend 1: How much is fixed and how much is variable?
Friend 2:15% variable. It is usually paid. So, 10.2 fixed and 1.8
variable. What about you?
Friend 1: I make 10 lakhs, but it is fixed. No variable pay though.
Friend 2: Anything tax friendly?
Friend 1: No. All based on payroll.

While the conversations are usually on salary, take home,

taxation benefit, and so on, benefits are never mentioned in the
discussion. While doing compensation analysis, it strikes you
when nearly 12% to 20% of employee payroll costs in India are in
the form of benefits.
Let us consider the benefit cost of an employee who gets a salary
of `10 lakhs. Let us assume that the person has a basic pay of
40% of the total CTC. A simple explanation of this would state:
Basic: 400,000
Allowances: 600,000
Total: 10,00,000
148 Winning on HR Analytics

Let us look at the benefits:

PF (@ 12% of basic): 48,000
Gratuity (15/26 of 1 month’s basic): 15,000
Telephone reimbursement (@2000 pm): 24,000
Tuition reimbursement: 75,000

This is an illustrative list. If you consider the amounts mentioned

alone, we get up to 16% of the cash salary.
Companies include defined benefits like PF to arrive at the
cost-to-company number. In reality, the CTC, when an employee
utilizes all benefits, is 8%–10% more than the CTC numbers.
For employees to shift their mindset from a cash compensation
base to a cash+benefits base, we need to clearly articulate a total
compensation statement that does not start with basic pay, but
with CTC.

Component Amount Percentage

CTC 1,175,000 100

Base pay 850,000 72

Variable pay 150,000 13

Benefits 175,000 15

It is sometimes very difficult to value benefits like employee

stock options and the like. However, it is recommended that as far
as possible, value the generic benefits and show them and use them
in discussions with employees. Using an analytical approach, HR
should establish the landed cost of each benefit instead of treating
benefits as a line item.

Portfolio Management of Benefits

While the basic salary may be 40% of the basic pay, it is a one
line item. On the other hand, benefits, even when they account
for 15% of the CTC, are comprised of 7–8 line items. This
 Optimizing Compensation and Benefits for High Performance 149

fragmentation ensures that no single benefit is too costly, but at the

same time makes it difficult for companies to assess their relative
importance.
To measure the impact of benefits, we need to consolidate the
following:

• Benefit,
• Target population,
• Number of beneficiaries,
• Percent workforce beneficiaries,
• Cost of benefit.

Target Beneficiary Cost/ Cost/

Benefit Target population number Percent Cost Beneficiary Employee

Library All 1,000 250 25% 2,50,000 1,000

employees 250

Extended Women 300 50 5% 25,00,000 50,000 2,500

maternity employees
leave

Food All 1,000 1,000 100% 2,50,00,000 25,000 25,000

coupons employees

Car lease Managers 200 120 12% 86,40,000 72,000 864

Tuition Individual 800 100 10% 50,00,000 50,000 5,000

reimburse- contributors
ment

Gymnasium All 1,000 100 10% 15,00,000 15,000 1,500

employees

Such an analysis helps us to take stock of how different

programs are doing. Food coupons are popular, as they cover all
employees. At the same time, benefits like library or gym, which
are open to all employees, are not used by as many employees as
they should.
Having such information helps us to identify targets for
rationalization. Total cost of benefits listed here is 4.3 crores per
annum. If we are to reduce this by 10% next year, what do we do?
In such a scenario, we look at the following options:
150 Winning on HR Analytics

1. Dropping a benefit altogether,

2. Offering it on a cost sharing basis. It can be 50:50, 75:25,
etc.,
3. Reducing the eligible population. For example, move the
company-leased car benefit to directors and above,
4. Effecting an across-the-board cut.

However, before effecting these cuts, one also has to compare the
benefits with their impact:

• Has the benefit utilization been increasing year on year? Is

the popularity of the benefit increasing?
• Is there any correlation between benefit utilization and
retention? Does a long-term benefit, like car leasing, help in
retention?
• Within the target population, some utilize the benefits and
some do not. Does the intent to use a benefit indicate intent
to stay for a longer tenure?

Armed with such data, we will be able to create a portfolio of

benefits as following:

Impact
Cost/ Utilization on
Benefit Coverage Utilization Employee trends retention

Benefit A All High Low Flat Nil

Benefit B Individual Low Medium Improving Positive

contributors

Benefit C Managers Medium High Declining Neutral

A portfolio of benefits, formed on the basis of utilization and

impact metrics, gives a view grounded in facts on evolving the
benefits strategy.
 Optimizing Compensation and Benefits for High Performance 151

Tailoring Variable Pay to Performance Based

on Data
As mentioned earlier, Sysco Corp., a $40 billion wholesale food
distributor based in Houston, found that its compensation system
for drivers—paying them by hours worked—did not provide as
much value to the organization as it could. According to company,
the model was not providing better customer satisfaction or
profitability. Instead, Sysco changed to a reward structure it
calls activity-based compensation. Drivers earn a base pay that is
supplemented with incentives for more deliveries, fewer mistakes,
and good safety records.
To be on the safe side, Sysco did not roll out the program
nationwide. It tested it in certain pilot markets first, and then tracked
the results of the operating company. Four metrics were targeted:
satisfaction level, retention, efficiency (delivering more cases in
less time), and delivery expense. Under the new compensation
structure, Sysco found that drivers were not only more efficient,
they were also more satisfied. The company’s retention rates for
drivers improved by 8% and expenses, as a percentage of sales,
went down. After quantifiable results showed the benefit of the
change, Sysco rolled out the program nationwide (CFO, 2006).
 10
Making the Transformation Possible

T he human resource management (HRM) function has

undergone many transformations since its inception as a
personnel and welfare department in the late 19th century. Every
transformation had added a new outlook or objective for the
HRM function and each time HR practitioners have been
successful in adapting to as well as adopting new transforma-
tional shifts. However, after the first decade of the 21st century,
HRM is again at the threshold of another transformative stage,
which is being labelled as people analytics, talent analytics, HR
analytics, etc.
The beginning of this transformation started with the use of data
in selection of baseball players who delivered stellar performance
resulting in multimillion returns for the team owners. This entire
baseball story got published as a book titled Moneyball in 2003
which subsequently caught eye of corporate world as it gave a key
message on talent management that need of the hour is to find the
talent who is being paid below average but has big potential. Thus
began the journey of data use in identifying talent and making data-
based talent decisions in the corporate world initiating the latest
transformative agenda for the HRM function.
However, this transformational demand on HR practitioners is
vastly different from earlier transformational changes as it poses
fundamental mind-set shift for HR practitioners. Unlike previous
 Making the Transformation Possible 153

transformations—this one is “subtractive” in nature than more on

the “additive” side, requiring HR professionals to create space
for a lot of unlearning. The challenge is further accentuated by
the nature of the work style of HR practitioners so far, which
has been more making talent or people management decisions
based on qualitative criteria and positioning HR professionals as
touchy-feely types, whereas HR analytics demands making talent
or people management decisions largely on the basis of data and
being more objective in decision making. Given the momentum
that HR analytics has gained in the last 4–5 years, there is hardly
any choice for HR professionals other than embracing this new
transformational shift and operationalizing it as a key discipline of
the mainstream HRM function.
The implementation of HR analytics requires HR professionals
to be fully aware of key “must haves” for successful execution.
Using a mix of McKinsey 7s Model, here is a list of key dimensions
(Figure 10.1) for operationalization of HR analytics and each will
be discussed in detail subsequently:

1. Skills
2. Systems (process and technology)
3. Structure
4. Support (leadership and culture)
5. Staff
6. Strategy (HR strategy)
7. Data presentation and communication

All of these dimensions are critical for the implementation of

HR analytics by any HRM function. Some salient features of the
model in Figure 10.1, before discussing each dimension in detail,
are:

• Support dimension, which includes culture and leadership

and hence this dimension permeates across all other
dimensions and thus underlines its importance by being on
the top of this model.
154 Winning on HR Analytics

Figure 10.1 Graphical Representation of HR Analytics Implementation

Schema

Source: Adapted from Watermen, Peters, and Phillips (1980).

• Data, like support, is another key dimension which is

sacrosanct for other dimensions and, hence, has crucial role
in implementation plan.

Now let us discuss in detail all the dimensions of Analytics

Implementation Schema as shown in Figure 10.1.

1. Skills: Implementation of HR analytics requires HR

professionals to possess these skills:
a) Business acumen: complete understanding of business
of the company/BU, how company/BU makes money,
and key value drivers of the business from the HR
perspective such as impact of employee wages, benefits,
productivity, and employee development investments,
and how these in turn impact business outcomes like
revenue and operating margins.
b) Statistical skills: basic knowledge of statistics to be able
to use and understand the meaning of various statistical
 Making the Transformation Possible 155

tools like measures of central tendency, dispersion,

distributions, correlations, regressions, probability, etc.
c) Research methodology skills: familiarity with the
research methodology field, like familiarity with
sample selection techniques, tool or instrument creation,
scientific approach to problem identification, and model/
framework selection.
d) Financial skills: basic understanding of key investment
evaluation methods such as cost–benefit ratio, RoI, net
present value (NPV), and internal rate of return (IRR).
e) Data presentation skills: these are required to make sense
of data along with statistical techniques, and include
graphs, plots, charts, and curves.
f) Data analysis: basic skills required include ability
to create a basic model, frame hypotheses, and test
and validate the same using data analysis tools and
techniques.
g) Data interpretation skills: this skill is different from data
analysis skill as interpretation requires the ability to read
patterns or interpret analysis results to draw insights and
actions.
h) HR domain skills: deep and wide understanding of
various HR disciplines such as workforce planning,
recruitment, performance management, development,
rewards, etc., and their interrelationships.

Patrick Coolen and Auke IJsselstein have created an “HR

competencies/skills wheel” required for successful implementation
of HR analytics (Figure10.2). This wheel also gives a glimpse of
the type of interface required with various other functions and
the kind of people needed to staff the HR analytics team, and is
covered next.

2. Staff: This dimension deals with “people” part of the

implementation model. An HR analytics team needs various
type of skills apart from those listed above and these other
skills include IT skills such as storing and retrieving data,
software use, advance statistical skills like algorithm and
156 Winning on HR Analytics

Figure 10.2 HR Capability/Competencies/Skills Wheel

Source: Reproduced with permission from Coolen and IJsselstein (2015).

model building, costing skills, business operations and

marketing skills, psychology and sociology skills to link
data with human and group dynamics, etc. So given the
varied nature of skills required by an effective HR analytics
team, here is suggested composition of team—organization
can borrow resources internally from other functions or hire
from outside depending on need:
a) HR analytics team leader: someone from a non-technical
background, though from HR domain, is ideal and has
excellent influencing and communication skills with
analytics expertise. Main role is to sell HR analytics to
business leaders and get their buy in. Additionally helps
 Making the Transformation Possible 157

in adding HR perspectives and context to interpretation

of data analysis results. Care needs to be taken while
selecting a person for this role as a wrong person without
appropriate gravitas and skills can kill the HR analytics
initiative.
b) Data analyst/statistician: someone who knows how to do
statistical analysis. He/she can be from HR or from other
functions like marketing or finance.
c) Data scientist: role of this person is to do big data crunching,
create models/algorithms, and interpret data results.
d) Data specialist: who maintains and ensures data is available
in the right format, and oversees the data governance process.
e) Other team roles/members required may include persons
with legal knowledge, finance, business operations, etc.
who can be drawn internally from other functions based on
requirements.
3. Structure: This dimension deals with issues like whether HR
analytics should be part of the HR function or not, and the
type of structure required. As far as the first is concerned, it
is ideal to host HR analytics within the HR function as it is
the HR head who is accountable to the CEO for HR decision
making and budget, etc., and it also helps in giving HR
language flavor to the analysis and results interpretation.
Regarding the type of the structure, an organization can
choose from the following formats based on its need, size,
and scale:

a) Centralized: here HR analytics group is at the corporate

level and rolls out the HR analytics framework and data
analysis results organization wide.
b) Decentralized: in this form each division or business
unit has its own HR analytics team for implementing HR
analytics.
c) Center of excellence (CoE): in this form there is an
expert team at the corporate/central level coupled with
an embedded HR analytics member for each function
or BU who is supported by the CoE team. CoE is more
158 Winning on HR Analytics

of an enabling setup for the implementation of HR

analytics across the organization. The CoE approach also
recognizes the fact that HR analytics needs a different
kind of skills and abilities than other CoEs, such as
learning and development, compensation, etc.
d) Functional: here, each function of HR, such as, learning
and development, workforce planning, or compensation
and benefits, etc., has its own analytics team or a person
to do analytics.
4. Systems: This dimension deals with the systems or processes
and technology assets required to produce and store data
for analysis—reports and dashboard types, generation
periodicity, tools or software assets or technology
requirements, data standards, protocols, templates, etc.—
to ensure consistent application of HR analytics across the
organization.
5. Strategy: Essentially, this piece covers HR strategy
encompassing HR analytics strategy and roadmap. Unless
HR analytics is made an integral part of the overall HR
strategy for the organization, utility of HR analytics
diminishes. Strategy dimension also includes clarity on HR
analytics strategy covering analytics roadmap, objectives,
metric selection, and their link to business outcomes in a
phased manner.
6. Support: This is one of the key dimensions whose importance
cuts across all other dimensions as shown in the model. It
has two sub dimensions, namely leadership and culture, and
each is elaborated below:
a) Leadership: means the C-level team of the organization
such as CEO, COO, CHRO, CFO, CTO, etc., and they being
on board for sponsoring implementation of HR analytics
by providing budgetary and other resources required for
its implementation. C-level support also ensures that data
which resides in other functions such as finance, operation,
marketing, etc., are made available to the HR analytics team
for analysis. Also more important is that the CHRO includes
HR analytics as a piece of his/her overall HR philosophy for
the organization to position HR analytics.
 Making the Transformation Possible 159

b) Culture: another key sub-dimension of support piece.

Culture of an organization has various nuances impacting
implementation of HR analytics. Basically organizations
either have a culture on a spectrum where one end is
the culture that believes in use of data for various types
of decision making including HR decision making,
and on the other end is culture which does not believe
in the use of data in decision making. Reason for lack
of cultural support for data primarily comes from high
transparency and accountability caused due to data-
based decision making. Also use of data in driving key
HR decisions such as recruitment, promotion, rewards,
development, etc., has the potential of upsetting the
political fabric of the organization. So one needs to be
sensitive to the prevailing culture of the organization and
how does that culture support or oppose the use of data
in HR decision making before deciding to implement
HR analytics.
7. Data: Last key dimension of the model is data, and its
importance for HR analytics is akin to blood running through
a human body. For this reason, data has to be treated as an
“asset” to be monetized through analytics. There are certain
key aspects related to data in the whole plan of HR analytics
implementation and these are covered below:
a) Nature of data: Typically, HR data is either unstructured
(text form) or structured (numerical), and good amount
of HR data is unstructured such as social media data,
resumes, survey comments, appraisal feedback, exit
surveys, training feedback, etc., requiring heavy coding
to make it machine usable. Other key challenges related
to data need to be taken care of and include:
i. Missing data: most of the time there is large amount
of data missing or not available affecting meaningful
analysis.
ii. Data reliability/integrity: primarily arises due to
lack of standardization of definitions, process
inconsistency, and methodology used for storing
160 Winning on HR Analytics

data. This can also impact quality of analysis and,

hence, needs to be sorted out before implementing
HR analytics.
iii. Data governance: deals with policy, processes,
responsibility, accountability, accessibility, legal
aspects, etc., for data storage, use, and communication
to appropriate stakeholders for decision making. It is
ideal to have a data governance group to take care of
these governance-related challenges.
b) Presentation: Another element related to data is
presentation of analysis results and here are some key
points to take care of this part:
i. Know the audience: tailor presentation of analy-
sis results according to the audience. For exam-
ple, the type of presentation for CEO will be
different from that for CFO, and similarly for other
business leaders also. Within HR, the type of presen-
tation will vary whether it is for CHRO or HR leader-
ship or unit HR managers or various CoEs of the HR
function. Also it is better to know analytics familiar-
ity of audience to make it easier to understand.
ii. Avoid cluttering: too many analyses at a single
place can lead to confusion or inability to grasp the
message.
iii. Focus on key points: identify the purpose of
presentation and restrict to that like whether it is
for employee engagement analysis or exit survey
findings, etc. Too many things at the same time can
lead to loss of focus.
iv. Insights and actions: more time spent on key insights
from analysis and actions for future generate more
interest and buy-in than discussing results or analysis
outputs.
c) Communication: It is another side of the coin while
one side is presentation covered above. It is being dealt
separately as communication of data to non-quantitative
audience like HR makes it more important. There are two
elements of communication requiring attention:
 Making the Transformation Possible 161

i. Storification: weaving a story around the data

results and insights makes attention, understand-
ing, and retention of analysis findings easier.
Story also helps in contextualizing the analysis
findings and, thus, gives organizational flavor,
increasing the receptivity. Also data analysis narr-
ated in the form of story works very well with audience
both at intellectual and emotional levels. Storifying
analysis results and insights is a skill which develops
overtime, and HR practitioners can learn it.
ii. Visualization: along with story, use of visuals adds
power to results and insights communication. In fact,
story and visuals make the results and insights self-
explanatory for audience to a large extent. But the
fact is that visualization is a poorly used and exploited
technique in HR data presentation and communication.
There are various methods of visualization depend-
ing upon the purpose of visualization such as sharing
of data, sharing of information, concept, or strategy.
Depending upon the purpose, visualization technique
will vary. Some of the ways to visualize data analysis
and results using different colors are:
a) Bars,
b) Charts—pie, line, radar, layer charts
c) Diagrams—Venn diagram, cycle
d) Boxplots
e) Graphs/plots of various types such as scatter
graph, curve graph, etc.
f) Tables
g) Matrices
h) Continuum
i) Cause–effect Chains
j) Maps—to highlight data findings for different
locations/geographies/units/functions, etc. Also
heat map
k) Network—with nodes
l) 3D diagram
162 Winning on HR Analytics

Executing Transformation—Rubber Hits the Road

After ensuring that various dimensions for operationalization
are in place, the next step is to start the execution. Given the vast-
ness of data and what analytics can do to data, it often becomes
challenging to decide where to begin. Many a times HR profes-
sional equipped with analytics tools start on fishing expedition
to find out something interesting from data. Analysis will always
throw up something interesting which may be of less business
value or no value at all. HR professionals often grapple with ques-
tions when rolling out an HR analytics program; some questions
are as follows:

• Where to start?
• Which metric is important to measure and analyze?
• Should focus be on attrition or employee engagement or
compensation?
• What type of data is needed?

Here is a step-wise suggested approach to kick-start HR analytics:

1. Begin with question: The “question” to be asked is “what is

the business challenge which has the HRM relevance?” Every
organization at any point of time faces business challenges
related to people management like attrition, employee
engagement, productivity, poor customer satisfaction, etc.
This question—“what are the key business challenges where
HR can impact?” should be asked to business leaders like
CEO and COO. Selection of a business challenge based
question to start HR Analytics will get buy in and support
of business leadership for doing HR Analytics. Support
and buy-in of leadership is lukewarm if analysis is done
of a question which is more close to the HR function like
performance appraisal rating trends, etc. The next step is
to select the top most challenge and take it forward. Some
sample questions can be:
 Making the Transformation Possible 163

a) Why is turnover high in some BUs?

b) What are the drivers of sales team productivity?
c) What is the impact of training on productivity?
d) Why is C-SAT low in some BUs?
e) What will be our talent gaps for the next year based on
attrition?
2. Create a model and hypothesis: After selecting one of the key
business challenges, the next step is to frame a hypothesis and
a model capturing that hypothesis. For example, hypothesis
can be that low employee engagement impacts customer satis-
faction, meaning that if an employee engagement is lower,
then it impacts a business outcome, like customer satisfaction.
A simple pictorial causal model between two variables—
employee engagement and customer satisfaction—can be
created to give visual power to the hypothesis. The next
step is to identify the top variables or drivers of employee
engagement. For identifying these variables, one way is to
use judgment or experience. Other valid method to avoid
errors is to read published literature on this topic and based
on organization context, a list of drivers of employee engage-
ment can be selected. Once narrowed down, these variables
can be added to the engagement variable in the causal model.
3. Select right measures/driver/variables and data: Once a list
of variables or measures of employee engagement has been
selected, then the next step is to identify which out of these
can be subjected to analysis based on data availability within
the organization. For example, studies may show that travel
or food quality is related to engagement, then the question
to be asked is whether data related to these measures (travel
distance/food quality) are available within the organization
to perform the analysis. If the answer is no, then these need
to be dropped. Also is there any past baseline or benchmark
data to compare trends and results overtime to draw robust
conclusions. If there is no past records of using such measures
or non-existing benchmarks, then it is belter to drop these, as
analysis results will be difficult to defend.
164 Winning on HR Analytics

4. Conducting analysis: once data has been collected, the next

step is to perform statistical analysis such as correlation
and regression using tools such as Excel, SPSS, R, SAS,
etc. Statistical analysis will help to validate or reject the
hypothesis and model. If it is rejected, then new variables
have to be selected to perform the next iteration of analysis
to check results based on statistical significance.
5. Interpretation of results: After statistical results have been
obtained, these have to be interpreted to make sense of
results. A simpler interpretation of results has to be written
keeping in mind the audiences with whom interpretations
will be shared.
6. Presentation of results/findings/insights and actions: The
final step is to share the outcome of analysis with the
stakeholders who have interest in getting this business
challenge. All the presentation techniques/tips outlined
earlier, such as using graphs/charts, creating a story around
results using HR language, highlighting key analysis
insights and action steps, have to be followed to make an
impactful presentation. It will be good foresight to prepare
a list of possible questions and keep answers ready to make
the presentation more professional.

Building an HR analytics capability in any organization is a

journey rather than an event. There are no right or best steps or best
processes to do this; however, there can be a guiding framework
based on the work done by other organizations successfully.
Each organization can tweak the framework based on its resource
availability and need of the hour. Over a period of time with
continuous improvements, HR analytics capability evolves.

Let us explore a few myths before completing the book.

People Analytics: Hype Versus Truth

There is a lot of hype around people analytics and in any tide
of hype, lot of “meaningless” boats get prominence while large
 Making the Transformation Possible 165

“meaningful” ones remain in the backdrop. This analogy aptly

captures the current picture in the world of people analytics. This
hype versus truth spectrum ranges from advocacy (Bernard Marr)1
for splitting HR into two functions of people analytics and people
administration to uberization of HR in the future. In between there
are lot of prophecies such as people analytics is a new fad marketed
by consultants or people analytics is the way forward for making
HR accountable like other business functions. Elsewhere, David
Green,2 in a recent post, wrote that it seems that HR has suddenly
discovered a gold pot in people analytics and every CHRO is
running after it! Here is our take on hype versus truth in people
analytics:

1. Hype (H): Data has value and is like an asset which can be
monetized for powerful insights, having impact on business
outcomes. Hence, data has to be treated as “strategic” in nature.
Truth (T): There is no doubt that data analysis can provide
powerful insights but truth is that organizations treat data as
“transactional” making it unfit for any kind of analysis and creating
a familiar situation of garbage in garbage out.

2. H: People analytics can help in identifying drivers of

organizational performance, making it easy for HR folks to align
HR initiatives and programs with those drivers to maximize the
organizational performance.
T: The reality is that we know lot about individual-level drivers of
performance such as pay, training, rewards, communication, etc.,
but how these individual drivers of performance are connected
to organizational performance is not clear as all individual
performance gets aggregated into a project and several projects
add up into a function or unit or account and, thereafter, sight of

1
 https://www.linkedin.com/pulse/20131118060732-64875646-why-we-no-
longer-need-hr-departments?goback=%2Enmp_*1_*1_*1_*1_*1_*1_*1_*1_
*1_*1&trk=nmp_rec_act_article_detail
2
https://www.linkedin.com/pulse/hr-analytics-new-gold-rush-david-green
166 Winning on HR Analytics

contribution gets lost. So we do not know organizational-level

individual performance drivers.

3. H: Big data in HR will help HR in becoming a truly “strategic

function” in the coming years.
T: HR function’s ability to use big data for HR will depend on
extracting “value” (V) from “variety” (V) of HR big data. These 2
Vs (of 4 Vs of big data) are central to success of people analytics.
Analyzing, interpreting, and making sense of data is not easy.

4. H: Buying a high-end technology solution marks the successful

initiation of people analytics function in the organization.
T: Technology at best can be an enabler of people analytics. There
is whole lot of ground work which needs to be done within HR
function/team and organization to harvest the benefits of people
analytics. One of the major reasons why people analytics does not
take off in many organizations is simply buying a highly marketed
technology solution and then assuming that the organization is
ready for people analytics. And also technology is expensive and
when it does not give promised results, disillusionment sets in.

5. H: Large datasets based on years of employee data are needed

for high-quality insights.
T: No doubt that large data are needed, but tiny datasets and
anecdotal evidence cannot be ignored and can be source of powerful
insights. Water cooler or corridor chats or food court murmurs, all
can be equally good data points for insights on effectiveness and
impact of HR programs and policies.

6. H: Models and algorithms bring perfection and rigor to the

practice of people analytics by bringing out patterns from large
amounts of employee data which is impossible for an average
person to visualize.
T: It is well known that even the best model and algorithm with the
best data can be imperfect as there are human limitations to include
everything while creating a perfect one for capturing everything of
 Making the Transformation Possible 167

a “whole man”3 that matters and that may get missed out in these,
and also, no perfect algorithm can give a perfect outcome.

7. H: People analytics will eliminate “human biases” as everything

becomes number based and numbers speak for themselves.4
T: People analytics has its own biases because it measures what
can be measured. However, people are people, and so far there
is no exact science to exactly measure and predict what and why
people do by uncovering all the biases and motives.

8. H: Job performance data can be converted into objective

measures making measurement simple and capturing each
employee’s contribution to the job and organization easy.
T: Over-objectification or over-simplification of the measurement
of performance creates the risk of missing the richness of what
makes that job special—or complex—or what makes each person’s
contribution unique. For example, an HR manager spending time
on a difficult employee or a client-facing manager who is soothing
down an angry customer, all are complex tasks not amenable to
easy measurement.

9. H: People analytics will help in eliminating errors in decisions

related to people management matters such as hiring, promotion,
potential, rewarding, and deployment.
T: It is true that errors will get reduced, but we run the risk of facing
unintended consequence of committing fundamental analytic
error’. In social psychology, we often commit fundamental
attribution error, meaning we attribute causes to intrinsic factors
(within individual) rather than to extrinsic factors (context or
situation). Hence, people analytics runs the risk of unfairly making
individuals who are not performing well a scapegoat based on data
analysis, while ignoring the weaknesses and constraints of the
system where individual is working. Truth is that system is always
powerful than an individual and most of the time overpowers even
the best talent in the world.

3
http://www.panarchy.org/whyte/organizationman.html
4
https://hbr.org/2013/04/the-hidden-biases-in-big-data
168 Winning on HR Analytics

10. H: People with HR domain background are best-suited for HR

analytics.
T: Deep HR domain knowledge surely adds value in making
sense of data patterns from “people” perspective but a broad-based
people analytics team comprising people with knowledge and
background ranging from business, financial, statistics (data sci-
entist), sociology, psychology, philosophy, IT, and data specialist
makes a deadly combination of a dream team for people analytics.

11. H: HR function moving from charting, dash-boarding, reporting

to predictive analytics signals the completion of installing people
analytics in the HR function.
T: Doing the analytics piece is half part of the people analytics
equation. True value and proof of pudding lies in converting people
analytics insights into meaningful actions for business outcomes,
which is often tough.
 Appendices

H R analytics is relatively a new area in HR function and the

adoption of HR analytics is picking up, albeit slowly. Majority
of companies with global employee footprint are still struggling
with creating a “global dashboard” of various HR metrics, while
adoption of predictive and prescriptive analytics is still lagging. A
study done by Meta Group and another by Bersin by Deloitte in
the year 2013, and adapted by John Macy (2015), found that in a
10-year period (2003–13) the adoption of strategic HR analytics,
such as prescriptive and predictive analytics, has increased from
4% to merely 14%, whereas the use of descriptive analytics and
reporting has decreased from 96% to 86% during the same period
(Table A.1). According to Macy, for any HR analytics to be of
strategic value, it should be 50% predictive and prescriptive and
rest reporting and descriptive. However, given the slow growth
rate, it may take decades to reach that level for any organization.
Clearly, many organizations are struggling to adopt HR analytics
and are at various stages of adoption of the HR analytics value
chain discussed elsewhere in the book. Apart from other factors
responsible for this slow adoption, one of the factors is related to
“taking the first step with clarity on where and how to get started.”
To address this challenge, from “practitioners” perspective, three
contributors who have led the implementation of HR analytics in
their respective organizations have distilled their experiences in
Appendices A, B, and C.
Though deployment of high-end analytics becomes a challenge
due to lack of tools and techniques, skills, and right data, yet even
a simple dataset which can be easily obtained can provide rich
insights about what is going on in the organization and how it
impacts key business parameters. Appendix D gives an illustration
170 Winning on HR Analytics

Table A.1 Workforce Information Management and Analytics (Maturity

Model—Comparison of Level 2013–2003)

of how using “employee identity number (EIN)” for data analysis

for a set time frame can provide rich insights.
One of the key areas of employee investment is training and
development and each year every organization spends millions of
dollars/rupees on employee development. Often getting approval
for this spend every year becomes a challenge for HR as leadership
keeps asking for RoI of spend in terms of hard numbers. Any HR
function without such hard data to showcase RoI of training and
development has to accept cuts in spends. A large automobile
group in India used data to track the benefits of training spends
 Appendices 171

at shop floor and successfully showcased its RoI to leadership as

explained in Appendix E.
The story is not complete without actual illustrations of talent
management analytics being used successfully in corporates.
Appendix F focuses on the power of analyzing text to arrive at out-
comes. Appendix G shares an example on using contextual search
to upscale a talent acquisition process. Appendix H articulates on
arriving at an RoI for talent retention. Appendix I is all about lever-
aging specific statistics for establishing engagement drivers.
Appendix J is one of the leading edge illustrations on the
emerging practice of predictive analytics. Appendix K signs off the
lineup with visual illustrations of the future: organization network
analysis.
The appendices, we hope, complement and support what has
been articulated through the book.

References
Meta Group.(2013). acquired by Gartner in 2014.
Bersin, J., O’Leonard, K., & Wang-Audia, W. (2013). High impact talent
analytics: Building a world class analytics and measurement function. Oakland,
CA: Bersin by Deloitte. Retrieved July 18, 2016, from http://marketing.bersin.
com/rs/bersin/images/hita100113sg.pdf
Macy, John. (2015, October). HR will never be strategic. Retrieved June 21, 2016,
from https://www.linkedin.com/pulse/hr-never-strategic-john-macy
172 Winning on HR Analytics

Appendix A

How to Get Started in HR Analytics

Having spoken at many conferences in the US, Canada, and the

UK, I’ve had the opportunity to speak to many HR professionals
from a variety of industries. When I ask people if they have started
the journey into HR analytics, many respond with a great desire to
get started but serve up one of the following five challenges as a
reason not to move forward:

1. Our data isn’t good enough.

2. We have no leadership support.
3. We’re busy working on other initiatives.
4. We have no budget.
5. We don’t have the skillsets.

Let’s take a brief look at each of these popular challenges.

Our Data Isn’t Good Enough

I’ve met quite a few companies who feel that the quality of their
data is insufficient to move forward with analytics. Most of these
are embarking on the implementation of a new HRIS system in the
hope of resolving their quality issues. These implementations are
estimated to take 18–36 months.
The companies that impress me the most are the ones that are
actively investigating their data quality to determine which data
they can use to start HR analytics and are forming a priority list
of what needs to be fixed in other data areas. This, I believe, is the
fastest route to getting value from analytics when you question
your data quality. After 25 years in analytics (in engineering,
supply chain, and HR), I can tell you that all of your data does not
have to be perfect to move forward. In fact, the reality is that your
data will never be perfect. There will always be reasons why you
have special “nuances” in your data to support business processes
that don’t quite match your technology processes.
 Appendices 173

We Have No Leadership Support

I’ve been there and I feel your pain, but this is never a reason to
not move forward. When it comes to analytics in HR, often the
leadership cannot envision what is possible. While 50% of new
CHROs are being selected from outside of HR in 2015 (and 30%
in 2014), the fact remains that most HR leaders in place today
came from traditional HR (aka non-analytical) backgrounds. It
will take specific examples of what can be accomplished with HR
analytics to help these leaders see the value. In speaking with HR
analytics leaders who have taken this approach, they report that
it takes about a year of these examples and much communication
before leaders start to envision other business applications for HR
analytics.
In December 2014, I conducted an informal survey and asked
HR analytics professionals around the globe what their biggest
challenge was in moving forward. The overwhelming response
was “leadership support.” It was not the support of HR leaders
though. It was gaining the support of leaders outside of HR. In
some of these cases, HR must realize that analytics pertaining to
the workforce may not be the biggest issue facing other leaders
in the company.

We’re Busy Working on Other Initiatives

…and you always will be. When I see that 85% of the companies
say that HR analytics is very important but 86% of the companies
report no HR analytics capabilities in HR, it tells me that many
companies are buried so far under in their day-to-day tactical work
that they have no time to consider anything strategic. In the current
world of rising workforce costs and headcount reductions, it is
difficult to find internal resources to support HR analytics. In these
cases, it might be well worth having a third-party expert start your
analytics initiative with a small pilot project that can be managed
externally.
174 Winning on HR Analytics

We Have No Budget
You don’t always need one. In the previous conference
presentations, I gave a specific example of how I, with a team of
volunteers across the globe, managed to produce a great deal of
analytical value for very little money. What’s the secret? Don’t
focus on what you don’t have; learn to leverage what you already
have. No one said that the first attempt at an HR analytics had to
look pretty! So, you don’t need to pay a vendor $500,000 to get the
value out of analytics.

We Don’t Have the Skillsets

I have helped several companies develop and execute their HR
analytics roadmaps, select technology, and communicate the
value of analytics to their employees. The key to remember in this
area is that it is more important to know which skillsets you need
and when. I have seen some companies hire a statistician in the
beginning of their journey only to realize that they will not need
this talent for at least a year. This risks having an under-challenged
employee that may not have the patience to wait for your journey to
catch up to his/her skills. Several smaller companies assessed their
needs and realized that they would not have enough of this type of
work to justify a full-time employee. In these cases, they outsource
their analytical projects.

So, Let’s Get Started

I would never dare claim that getting started in HR analytics is
easy. It is admittedly challenging in that it requires a great deal
of thought, assessment, and the persistence to change traditional
mindsets. It is truly an uphill battle, but if you thrive on challenge, it
is greatly rewarding. Here are a few suggestions to get you started:

1. Start small and find an area of the company that is feeling

some sort of pain related to the workforce. They will be
thankful for any help you can provide and they will be your
biggest supporter if you can show results.
 Appendices 175

2. Assess the current state of your HR analytics program. An

Internet search on “HR analytics road maps” will yield
several expert views on the progression of HR analytics
capabilities. Compare your current capabilities to those
sample maps and determine which new components you
would like to add to your company’s capabilities in the next
year or two.
3. Focus on specific business questions. If you do not know the
specific questions you are trying to answer or the specific
problem you are trying to investigate, you risk wasting a
lot of time and money on something the company doesn’t
value.
4. Don’t buy HR analytics technology until you know how
you plan to use it. These investments are very expensive so
it is best to conduct several analytics projects without this
technology to get an idea on what your company really needs.
Only then should you search the market for technology that
matches your needs.
5. Analyze your data in context. Numbers are great and many
people are skilled at “slicing and dicing data.” But, what
does that data mean to the business area or the problem you
are studying? This is where HR analytics experts need to
partner with experts in the specific business area to be able
to interpret what the data is really telling you. For example,
what are the possible root cause reasons for the movement
of metrics in the data? People working in that area every day
are best suited to provide this insight.
6. Expect the unexpected. Your first few projects will take
longer than you anticipate as they reveal multiple categories
of challenges along the way that you did not anticipate.
While painful, these will also be the projects from which
you learn the most about how to do HR analytics within your
own company.
7. Seek outside help at the beginning of your journey. Many
people have been down this path before you who can offer
advice on what went well in their HR analytics journey and
what did not. Learn from the wisdom of others.
176 Winning on HR Analytics

Tracey Smith is one of the “Top 50 Global Influencers in HR

Analytics” and presents at conferences in the US, Canada, and
the UK. She holds degrees in Mathematics, Engineering, and
Business and has over 25 years of experience in the areas of
Human Resources, Supply Chain, and Engineering. Tracey is the
author of “HR Analytics: The What, Why and How” and “Strategic
Workforce Planning: Guidance & Back-up Plans.” She is also the
owner and editor of “NI Magazine,” an e-magazine dedicated to
the HR analytics and workforce planning communities.
(Source: www.numericalinsights.com (accessed on 22 August, 2016).)
 Appendices 177

Appendix B

Seven Deadly Sins of HR Analytics Initiatives

The right HR leader can play a vital role in enabling workforce

analytics initiatives. Effective sponsorship will ensure that
an analytics initiative has the support required to succeed, the
resources—both people and technology—to be impactful, the
support to create and communicate a clear, compelling vision,
and the leeway to focus its attention on key business challenges
or issues.
Conversely, the wrong HR leader can help to kill a talent
analytics initiative before its first breath. With the best of
intentions, a misguided leader will—intentionally or not—
engage in behaviors that virtually guarantee the failure (or
significantly impeded the traction) of future workforce analytics
initiatives. In some cases, these errors are committed with
the best of intentions. In other cases, good intentions are not the
issue, but rather ignorance, naivety, or hubris (or a combination
of the three) plays a role.
If you, as a HR leader, want to really support your organization’s
HR analytics efforts, there are specific things—“sins,” if you
will—you should seek to avoid. Granted, “sin” may be a strong
word, but hey, it got your attention and may have influenced your
desire to read this! However, these are dynamics that—if you can
avoid them—help set up your workforce analytics initiatives for
success.
So, in no particular order, here are seven dynamics that can
surely make analytics initiatives more difficult (or set the stage for
failure at inception):

1. “Burger King”—having it your way (and changing “your

way” at a whim)
2. “Teacher’s pet”—solving for your pet issue
3. “Penny wise and…”—refusing to invest strategically
4. “Headcount neutral”—requiring cuts in other areas to fund
analytics
178 Winning on HR Analytics

5. “Other duties as assigned”—making analytics a leader’s

part-time role
6. “Dummying down”—assigning analytics to report to a “less
analytically inclined” leader
7. “Know it all”—believing you are the person to “sell”
analytics

“Burger King”—Having It Your Way

This is one of the most common issues—a senior leader decides

that analytics is going to support their “vision”—which (unfortu-
nately) is often no more than a better way to see my “dashboard”
or “scorecard” or to demonstrate that, “I am doing what I said I
have been doing” (or, actually, “others are doing what I said I was
doing”). It’s not about measuring the most relevant people-focused
business drivers, but rather providing a more technologically
sophisticated “veneer” on their “pet” metric, report, scorecard, or
dashboard.
Compounding this is the leader who cannot fix on a given
objective or set of objectives, but instead change their “order” as
one might change what they order from one trip to Burger King to
another. They start out with no vision regarding what they want
and then, in the absence of any clear vision, their interests shift
from one metric to the next ad nauseam. Those who are lovers
of the “Burger King” approach to analytics do not want insights
necessarily; as much as anything, they are good with whatever
they have a whim to measure at any given time. Unfortunately,
in attempting to simply re-create what they already have, they
stifle the ability of their HR analytics’ team to deliver strategically
significant insights.

“Teacher’s Pet”—Solving for Your Own Pet Issues

“Teacher’s Pet” is quite similar to the preceding characteristic.

The difference is that it is focused on furthering a specific agenda.
Most commonly, you see this in the HR leader who supports
 Appendices 179

analytics in order to “prove” or “validate” that their favorite pet

project is yielding the promised results.
HR analytics can also be hijacked by the leader who is seeking to
show others where there are deficiencies in the organization (rarely
or never their own) or in other leaders. In this case, it is all about
“analytics by vendetta,” with the focus being on shining the light
on a colleague’s people issues, not only on providing compelling,
enterprise-relevant insights about their people.
Either approach (“proving” the value of “pet” projects or
“torpedoing” an adversary’s project) is detrimental to the efforts
of your workforce analytics initiative. How? Nothing kills the
credibility of workforce analytics faster than it being perceived as
an “arm” of a self-promoting leader.

“Penny Wise and…”—Refusing to Invest Strategically

This is most commonly observed with leaders who attempt to

initiate analytics programs by making their team use inadequate
technology. Itis not that the workforce analytics initiatives
necessarily require significant investments. However, there is—in
most cases—incremental investments, whether they are people or
technology-related.
This is often most apparent in organizations where a senior HR
leader will not support any incremental investments, requiring the
workforce analytics leader to beg, borrow, or—god forbid—steal
the resources necessary. The bottom line is that they do not have
confidence in the value of the initiative and, hence, do not want to
risk any of their own financial or political capital.

“Headcount Neutral”—Requiring Cuts in Other Areas to

Fund Analytics

This is very similar to “Penny wise”; the primary difference being

that the leaders who embrace “headcount neutrality,” deliberately
or otherwise, put analytics in a “no-win” situation, competing for
limited resources (people, dollars, technology, etc.) with their
180 Winning on HR Analytics

peers (who are, like HR analytics, competing for limited resources

and their own survival).
These leaders place their HR analytics initiatives in sort of a
“Hobson’s choice.” Named after Thomas Hobson (a 16th-century
British livery stable owner known for his practice of offering
customers the choice of either taking the horse in the stall nearest to
the door— as a means of rotating the use of the horse—or none at
all), a “Hobson’s choice” is a free choice in which only one option
is offered. As a person may refuse to take that option, the choice is
therefore between taking the option or not. In plain English, itis a
form of “take it or leave it.”
Leaders who use this approach may, erroneously, believe
that they are being good stewards of their organization’s budget.
However, what they are, in fact, doing is pitting their HR analytics
initiatives against all others seeking the same source of funding.

“Other Duties as Assigned”—Making Analytics a Leader’s

Part-time Role

In many cases, this is also related to a dynamic described above—

the leader does not appreciate the potential value of their analytics
investment (and are, in effect, looking to get something for
nothing). The problem, in a nutshell, is that what they often get is
just nothing. Or they get very little in the way of progress because
they are not willing to allow a team member to focus their efforts
on the opportunity.
Granted, this is not always bad, per se. In some cases, this
approach can be a great way to start, in part, because it makes
the RoI so attractive. Also, for the part-time analytics leader, he/
she has very little pressure to produce , because the expectation to
deliver is so low.
However, for those leaders who are truly committed to HR
analytics and are desirous of timely, impactful results, this is rarely
the best way to go but many will still choose to do so.
 Appendices 181

“Dumbing Down”—Assigning Analytics to Report to a

“Less Analytically Inclined” Leader

Most commonly in these cases, HR analytics is assigned to report

to a vice president of talent, organizational effectiveness, or
some other function. In theory, it makes sense on some level; the
assumption is that HR analytics is a “centers of expertise,” as is
staffing, compensation, benefits, and other functions.
Unfortunately, the assumption that HR analytics is just another
“CoE” is flawed. It is a capability possessed by very few HR
organizations (unlike staffing, compensation, benefits, etc.) and
with good reason, it requires a specialized set of knowledge, skills,
and abilities.
Trust me, it is a rare “less analytically inclined” HR leader who
can effectively advocate for and support an HR analytics initiative
of any size, scale, or complexity. I have tried it. I would never do it
again. Life is too short. I will leave it at that.

“Know It All”—Believing You are the Person to “Sell” Analytics

An extension of the preceding “sin,” “know it all” HR leaders

(who are often also “less analytically inclined”) take upon
themselves the role of “selling” HR analytics to others within
the HR organization and the company as a whole. Unfortunately,
lacking a background in analytics, scientific methodology, or
research methods, it is the rare non-analytics HR leader who is
equipped to do so.
When this is paired with a “less analytically inclined” leader
who also is insecure in allowing members of the HR analytics team
to represent the work done by the analytics team to other leaders,
these issues only become more apparent and put the efforts of the
analytics team at risk, as they are being voiced by someone who
cannot begin to form complete sentences with respect to the work
being done, not to mention explaining the nuances of the projects
undertaken.
182 Winning on HR Analytics

Prologue
Hopefully, I have been able to shed light on some of the “sins”
prevalent in the field of HR analytics. If, in looking in the mirror,
you see a “sinner” staring back at you, there is only one true option:
Repent! Recognize the error of your ways, seek to make right your
wrongs, and put your feet on the path leading to what is good
and proper. HR analytics initiatives, not plagued by these “sins,”
can provide strategically significant and sustainable competitive
advantage. Do not allow yourself to become (or continue to be)
ensnared in the errors of these sins.
Now, go forth and sin no more.
(Source: Mark Berry, Vice President, HR, CGB Enterprises Inc., Covington, LA, US.)
 Appendices 183

Appendix C

Starting with Workforce Analytics? Five considerations

before Taking the Leap
Interview with Bill Roberts, for Harvard Business Review
Analytical Services by Stela Lupushor.

If you are starting on your journey to use analytics in the context

of HR—congratulations! The timing could not be better. With
analytics in HR becoming a popular topic, there are more skilled
professionals who are interested in switching their specialization
and who bring a wealth of knowledge and experiences to HR. All
this is good, but where do you start?

1. Hire “top guns”. This is a new area for HR and it requires

a very different skillset than a typical HR professional will
have. Consider individuals with degrees and experience in
statistics, data science, computer science, I/O psychology,
consulting. They can learn HR processes much faster than
training a traditional HR professional on technical skills. Be
prepared, they might not come cheap! This is an area where
the “war for analytics talent” has been won by the “talent.”
But giving HR the ability to consult the business with facts
and analytics-driven insights is priceless!
Start with the business questions. There will be plenty of
things in HR to focus analytics projects on, but the biggest
value will come from linking the data about your workforce
to business outcomes. What are the characteristics of
employees that bring in a disproportionate amount of
revenue, or have the most impact on customer satisfactions
scores? How can you focus hiring and development activities
on those characteristics? How can you tweak learning
programs to bring employees to a productive level faster?
184 Winning on HR Analytics

2. Ruthlessly pursue the right problems. An even more

important question is, “Are we solving the right problem?”
For example, predicting attrition risk is one of those
perennial issues that many mature analytics functions
attempt to solve. There are many predictive models that
have been created over past decades (yes, decades). We have
more sophisticated tools to perform analyses. We now have
more data, a whole lot more of it—structured, unstructured,
internal, and external. We can quickly assess the supply and
demand for specific skills, which location—down to a city,
where those skills are in great demand, which profiles our
competitors are looking for, what are the characteristics
of those who might be at risk, etc. We are talking beyond
typical segmentation and getting to a very granular analysis
of online behavior. But guess what? Little has changed. All
this sophistication still does not change the fact that people
leave. People will always leave. The question is whether one
needs to spend time trying to prevent the inevitable or think
about the problem through a different lens. The questions
that can be asked of the data instead are more about how to
get the most out of the employee while he/she is with the
company.
3. Data perfection might be optional. Being on the same
page with stakeholders—mandatory! There is nothing
more damaging to an analytics team’s reputation, especially
at the early stages of building the analytical mind-set
in HR, than poor data quality. People will discover that
“Bob is missing from the report” and discredit the whole
analysis as a result. With the increasing amount of data
there will be more opportunities for error. Is 80% accuracy
sufficient to make the right decision? Would you get more
value from your analytics team having time to run another
analysis versus going on a “finding Bob” expedition? The
surest way to mitigate the concerns, however, is to agree
with the stakeholders very early on about the expectations
for accuracy, the analysis methodology, the type and
completeness of data, the structure of the report, and types
 Appendices 185

of recommendations that will come out of it. It might take

more time up-front. But it will possibly be the only thing that
prevents the analysis from ending up in the “discard pile.”
4. Be a humanizing force. HR professionals have a huge
impact on the working environment of many. Analytics
gives great insights and power. Take a step back and think
of its impact. Put yourself in the shoes of those employees.
Would the change make you personally feel comfortable?
Would it allow you to realize your full potential? Would
it inspire you? If you are hesitating to answer, you might
not be solving the right problem. There is a human being
with dreams and aspirations, behind every number in your
analysis—a human, a social creature, who wants to belong,
to connect, who wants someone to care, and to be treated
with respect and dignity. Do your job such that you create
an environment where you and other people can thrive. And
use analytics for good; use it to tell the story of that human
being, use it to humanize the workplace.

I would like to end with this thought. The rapid changes, onslaught
of data, and analytic capabilities necessitate strong leadership,
conviction, boldness to try non-conventional methods, to take
risks, and ask the right questions. It also means that sometimes you
do not know whether the path you are taking is the right one and the
only way to know is to give it a try. Have the courage to try and do
not forget to share the scars, bruises, and everything else you have
learned with all of us!
(Source: Stela Lupushor, Vice President, HR Planning & Analytics, Fidelity Investments,
NY, US.)
186 Winning on HR Analytics

Appendix D
SMALL DATA CAN BE AS POWERFUL AS BIG DATA
As the Shakespearian quote goes “What’s in a name?,” one needs
to look beyond name to get the meaning or essence of the object.
Same is the case with EIN, which is equivalent of a personal name
record (PNR) in the organizational context providing link to all
kinds of demographic and work history data of the employee. That
seems to be a simplistic way of looking at EIN without attempting
to understand the powerful insights this number can provide into a
working of an organization.
EIN is allocated to a full-time employee on joining upon
the organization and in some organizations, all types of employees,
such as full-time equivalent (FTE), contract, contingent, tempo-
rary, etc., have same continuous series number, whereas in some
organizations EIN for FTEs is different than that for non-FTE
employees. Irrespective of how it is being deployed, use of people
analytics approach can throw up interesting insights hidden behind
this number. Like any people analytic approach, EIN analysis can
be applied to a specific organization to get required insights about
workings of the organization. EIN as a tool is like EBITDA or EPS
analysis which, if done properly, can provide a lot of information
about any company working.
Let us take a case of a company to illustrate this. An employee,
John joins in company X in June 2004 and is given an EIN as
19450. Company X was founded in 1995 and operates in the IT
services sector. IT services sector typically has an attrition rate of
15%–20% per annum and business growth rate ranging from 10%
to 20%. After few weeks of joining, John found that total employee
headcount including all type of employees is 4,600. John worked
in the company for 8 years and left in April 2012. At the time of
leaving John found that latest EIN is 52157 and total headcount is
12,350. A quick analysis of EIN data for company X from 1995 to
2012 shows that:

1. From 1995 to 2004, company had hired 19,000+ on its rolls

during the 9-years period and the actual headcount in June
2004 was 4,600.
 Appendices 187

2. From 2004 to 2012, company had hired 32,000+ employees

during the 8-years period and the actual headcount in April
2012 was 12,000+.
3. And company, since its inception in 1995, had hired 52,000+
employees till 2012 but the actual headcount in April 2012
was 12,000+.

A quick interpretation of the above data shows that:

1. From 1995 to 2004, headcount churn was close to 5 times in

9 years.
2. From 2004 to 2012, churn was 2.5 times in 8 years.
3. And from inception till 2012, churn was 4 times in 17 years.

If the company A had grown within the growth rate range of

10%–20% and had an attrition rate in the range of 15%–20%, it
would had have experienced a homeostatic state on the headcount
front.
Nevertheless, using the data analytics approach, what insights
does these data provide about the business and growth of company,
people management approach of the company, culture, and more
importantly a peep into the operating philosophy of the senior
management? Let us try to derive some insights from this analysis
by looking at the impact on business outcomes:

1. Business and business growth quality: Data analysis

points out that
• Company has witnessed fast growth at different time
periods in 17 years of its existence.
• Growth is not a stable one and has periods of high growth
followed by a sharp drop in growth. This is evidenced by
high volume hiring of 32,000 during the period 2004–12
whereas the actual survival is close to one-third.
• That company lacks a focused business strategy with
clarity on what business it is in and what business it is not
in, similar to an “Indian thalli” (a large plate served with
everything available to eat!) approach which has spread
of all dishes, leaving the consumer confused. It points out
188 Winning on HR Analytics

that company is unable to figure out where to invest and

grow and where to get out.
• Company has been unable to develop core competence
areas as a differentiator from similar companies in
the industry as this would have enabled to maintain a
steady growth in core areas without witnessing too
much turbulence in employee headcount while allow-
ing it, at the same time, elbow room to explore emerging
areas.
2. Culture of the company: In the absence of data related to
how many employees have the longest tenure in the company
since its inception, it can still be concluded that culture will
always be in a flux with the absence of a true culture acting as
a glue for creating identity of the company. As churn ranges
from a multiple of 2.5 to 5 over a period of 17 years, one can
easily imagine challenges related to process implementation
and continuity as well as creating a cultural identity for the
company.
3. Customer satisfaction: With such a high employee churn
rate and low tenure rates, customer service will suffer as
service level agreement (SLA) is impacted due to con-
stant deployment of new employees and thereby inability
to deliver consistent quality service to customer. As a con-
sequence, one can safely assume that company might be
experiencing high customer churn also.
4. Senior management quality: As the data shows, 12,000
is the headcount in 2012, whereas 52,000 were hired since
1995, so it implies that company has witnessed complete
replacement of employees four times in its life cycle so
far barring few (assuming 5–10%) continuing over longer
periods beyond 10 years tenure.
a) This points to senior management’s preference for mirror
images of self at different levels that can have a longer
tenure.
b) Also it highlights deficiency of sound business and
operational knowledge coupled with lack of leadership
 Appendices 189

skills in senior management as evidenced by the inability

to build a sustainable and lasting organization (or
institution).
5. People management: As in any IT services sector company,
people are the key drivers and differentiators of business,
and insights based on company X’s data analysis shows that:
a) Company believes in just-in-time hiring for growth as
headcount maintained is linked to what business can
sustain.
b) Just-in-time hiring raises the overall employee costs as
lateral hiring is always costlier than developing from
entry level upwards, thereby providing less elbow room
for employee investment or innovation.
c) Company has pure “task-based” (or project based in IT
parlance) approach towards talent rather than “people”
or “individual” based. What this means is that as soon
as the project or task is over, an employee loses its value
unless he/she can be deployed for a similar task, if avail-
able. People- or individual-based approaches signify
that organization values an individual beyond the skills
for current task or project and is willing to help the indi-
vidual to grow and develop over a longer period by skill
development.
d) Employee development and growth is purely a tick-mark
approach as evidenced by the lack of company’s interest
to maintain a pool of people for projected or anticipated
growth. Clearly, talent management will be restricted to
task or work at hand with no planning for future.
e) HR will be heavily transaction-focused as large volumes
of hiring and onboarding and exit management will
consume significant amount of HR team.
f) Also just-in-time hiring has another unintended conse-
quence of pulling resources from ongoing projects for
deployment to new projects to prevent the revenue loss
till the time new hire joins, which creates adversarial
relationship between HR and line managers as the latter
sees HR prowling for people for new projects.
190 Winning on HR Analytics

6. Employer Brand: More than 40,000 people have passed

in and out of the company doors. Thus, company has huge
alumni outside at any point of time than the actual employee
strength. Now these alumni might have worked for different
periods of time while in the company, more likely joined as
laterals, and had good and bad experiences at the company.
Assuming only half of alumni talks good about the company
while the other half shares negative things, still the number
is huge for any company to sustain and build its employer
brand to keep attracting quality talent.

Above narrative based on simple analysis of EIN data and looking

beyond what EIN hides shows that data analytics, even if done with
very small and simple dataset, can be source of powerful insights
about what is happening inside the company. Typical HR person
need not to apply sophisticated algorithms to people analytics
for getting powerful insights. And if this EIN data is supplemented
with more data, such as quality of hire, quality of performance
ratings, manager and senior management churn rates, actual attrition
rates, actual business growth rates and customer satisfaction, etc.,
more powerful insights can emerge on the state of affairs of a
company.
(Source: Kuldeep Singh.)
 Appendices 191

Appendix E

Establishing RoI for Training Investments

It was April 2009. The global financial crisis had hit the world and
countries across the globe with no exception were in some way or
the other impacted. The mood in the corporate world was rather
somber.
The organization ABC Private Limited is a manufacturing
organization of renown. It manufactures auto components and
has its sole manufacturing location situated in Central India. The
company supplies its products to both original equipments (OEs)
and the retail space (aftermarket) in the Indian domestic market.
The company had no immediate plans to explore opportunities in
the export markets.
The company was facing tough competition in the market place
and was not able to grow as much as it would have desired to.
The global recession had its impact and was taking a toll with
customer orders on the wane. The capacities were lying idle and
the company’s huge investments in capacity expansion (of over
100 crores), with a presumption that customer orders would fully
utilize plant capacities, was not happening as had been anticipated
whilst planning the capex investments.
Also compounding this were its droping productivity and
quality levels as well as spiraling fixed costs due to addition of
manpower to man the additional capacities. Quality complaints
from the customers were also on the rise. The senior management
was truly a worried lot. Hitherto, the company had a large market
share in the OE space and was garnering close to 35% of share
of business (SOB) and had a healthy market share of 15% in the
aftermarket space as well, with an overall compound annual growth
rate of 16% over the last 5 years (FY03–FY07).
The market indicators were pointing to the organization losing
its OE business as well as competition eating into its aftermarket
space. The SOB in the OE space had shrunk to 31% and the after-
market SOB to 11%. This was a huge dip and had the organization
worrying and scramble for some solutions to fix this. Some key
192 Winning on HR Analytics

personnel from middle to senior levels too had put in their papers
compounding the problem in hand.
The organization had started as a family-owned small scale
industry in 1978 and had, over the years, grown to a mid-sized
organization with a turnover of over `1400 crores. The company
had professionalized over the years and the CEO and other senior
team members were professionals hired from companies of repute.
The promoters had decided to step back from day-to-day
management and had completely empowered the CEO. The CEO
in fact got the full support of the promoters and the board as well
since his induction into the organization 5 years ago.
The board met with the senior leadership team to discuss
possible options to mitigate the problems that the company was
facing. It was decided after long parleys that the need of the hour is
to seek external consulting support.
The company decided to take this ahead and sought the support
of a renowned global consulting firm to advice on its business
strategy. The consulting firm did a detailed study of over 12 weeks
and came out with its findings and recommendations. The key
recommendations were to enhance the visibility through brand
promotion, relook at its channel management strategy, its portfolio
of offerings and supply chain management (SCM) strategy (to
contain costs), enhance its focus on safety, and most importantly
beef up skills and competencies of its human capital across all
levels of the organization.
To elaborate, skills were more from an execution/work
standpoint, whilst competencies were supervisory/managerial in
nature.
Based on the advice of the strategy firm, a formal mapping of
the key roles and evaluation of skill and competency levels were
undertaken by a leading HR consulting organization. It was clear
from the study that there was a yawning gap between desired level
of skills and competencies (based on the strategic business intent)
and the actual ground reality.
The CEO was apprised about the matter and decided to meet with
his senior management team to address this skill and competency
gap issue, which was seen as crucial to the very existence and
growth of the organization. The CEO committed to invest at least
 Appendices 193

3% of the annual turnover in training each year and gave this

assurance to the senior team when they met.
The senior management team met to discuss the execution of
modus operandi and next steps.
The CHRO was anointed the project manager of the project
aimed at driving the skills and competencies enhancement. This
was christened “Parivartan,” (transformation) since it was agreed
and understood that this is not a skills development project but a
larger project of change, requiring a paradigm shift in the very
thought process of learning and application of knowledge. This
also included ushering in knowledge management practices and
tapping into the tacit knowledge of members gained over the years.
It will be pertinent to note that attrition was hovering around 8%
on an average over the last 5 years. Also, many employees had
been with the organization for long and the average tenure of an
employee was about 10 years.
With the flagging off of Parivartan, the die was cast for a
multipronged approach to address all levels of the organization
from a training and development perspective.
The organizational leadership decided that a synchronous train-
ing strategy and not a sequential one, is the need of the hour. There
was also no time to lose and it was important to act. Identified
senior leadership members were nominated for general manage-
ment programs conducted by management institutes of repute in
India. In fact, two high flyers, which were in the succession plan for
the CEO’s slot, were nominated to the Ivy League universities in
the US for a long development program spanning three months.
Some select senior members were also mentored by life
coaches. The middle management team members were nominated
for specific functional, developmental, and skills upgradation pro-
grams. The junior management team members were provided func-
tional training as well as nominated for a development program
aimed at honing their supervisory and decision-making skills. An
e-learning portal was created and technical, functional, and man-
agement modules were made available to all employees across,
in English, Hindi, and vernacular. This helped employees learn
at leisure and use these learnings to do better at their work place.
194 Winning on HR Analytics

Hitherto, the training was addressed more as a by-product of

business outcomes. The paradigm shift now was that the training
was no longer an activity that was just nice to have, but an important
business driver that had to be invested in and nurtured assiduously.

The investments were steadily but surely beginning to pay off.

FY 13 saw a sea change in the approach as well as the attitudes

of the members all across. The buoyancy was palpable even when
the business plans for the fiscal were being drawn up. From an
approach of “can we do it” the paradigm shifted to “we will do it.”
The organization created a slogan “Hum Honge Kamyaab” (we
will succeed) and this created a lot of excitement and an adrenalin
rush across the rank and file.
The investments in these training interventions over the last few
years had paid rich dividends. The results of FY 14 were speaking
for themselves. The top-line grew by almost 22%. The cost of
employing expense to sales hovered steadily around 6%–7% over
the last 3 years—a remarkable achievement. The sales per head
count almost doubled. The OE market share and the AM SOB
improved by 3% in both the segments.
From an operational perspective, the focus was on enhancing
the profitability through prudent energy management, ergonomic
plant layout design, easing of material flow, low-cost automation,
and reduction in rejections through better adherence to defined
manufacturing and quality processes. A thrust was given to drive
total quality management (TQM), total productive maintenance
(TPM), and lean manufacturing and an organizational productivity
head was nominated to champion these initiatives with a committed
cross-functional team, who were christened “change evangelists”
to anchor and evangelize project Parivartan across the organization.
Safety was given paramount importance and safety training was
prerequisite every six months for all employees CEO downwards.
A formal safety certificate was required for the employee to work
on the shop floor and this was to be renewed every six months after
being duly certified by the safety head. A leading multinational
firm was co-opted to advice on safety and work closely with the
 Appendices 195

safety and operations team. The point to note is that safety was
made an integral part of the organizational working credo across
all functions (not just shop floor operations). Safety was not just a
staff function playing an advisory role but an important strategic
arm of the organization.
This thrust on safety ensured nil accidents and the company saw,
very minimal near misses. The impact was that man-days lost on
account of accidents came down by 25% over a period of 3 years.
The organization, therefore, was saved considerably as there were
no man-days lost due to absenteeism. The morale and motivation
of the employees also took an upswing, as they were convinced
that they were in a safe working environment and no harm would
befall them or their families.
Through education and training of employees, on the need
and benefits of energy conservation and the use of wind/solar
power (where feasible), the specific energy consumption came
down by about 5% YOY and this helped the company save about
`12 crores per annum, a considerable sum. The impetus on TQM,
TPM, and lean manufacturing saw rejection levels come down by
about 10% over a 3-years period. Through prudent SCM processes
and quality processes, the company was able to save about 2% in
its material costs over a 3-years period and this amounted to about
8 crore per annum of savings. The savings, in a sense, more than
compensated for the investment on training, and the greater intan-
gible benefits were rise in employee morale, motivation, produc-
tivity, and overall organizational health.
The process of change was through constant education, employee
involvement, using cross-functional teams to solve organizational
issues together, suggestion scheme rewards, creation of a TQM
club, and quality circles. The key was that these were driven
by the employees themselves and were surely the cornerstone
for the success of these interventions. Constant recognition and
commensurate rewards helped shore up employee morale and
motivation levels. A pointer to this is that the overall employee
engagement index went up from 65 in 2010 to 74 in 2014.
Given the flurry of activities on the people development and
training front, the attrition of mid-to-senior level employees too
196 Winning on HR Analytics

was not there and this helped create a stable working/strategic

team in the organization. The reputation of the organization as
an employer of choice also helped ABC attract lateral talent and
added power to the talent pool of the organization.
It is also important to mention that the number of days per
employee went up from 3 days a year to 6 days a year, a 100%
increase over a 24-month period. The Kirkpatrick Model was used
to measure training effectiveness using the four levels, namely
reaction, learning, behavior, and results. There was a perceptible
change that was visible in the behavior and this manifested itself
through visible improvement outcomes in terms of the results.
The success and turnaround of this organization from a
training and development perspective have induced many other
manufacturing outfits in the near vicinity and also across the
country, to follow suit. The belief that human resources are
a differentiator, that no product or technology can ever be, has
finally dawned. The power of human resources of ABC Company
has positioned it on a strong pedestal and given this company an
edge to be a market leader and also be an employer of choice for
aspiring talent.
As the CEO Remarked, “The proof of the pudding lies in its
eating.”
The overall business outcomes were ample testimony to the
benefits accrued through a concerted effort on the training and
development on all fronts. The investment in terms of man-days
on training doubled, there was a behavioral shift towards looking at
the future with confidence, business outcomes in terms of enhanced
top-line and bottom-line growth were clearly visible, safety indices
were showing a surge for the better, and most importantly employee
engagement indices had improved, indicating the fact that the
overall health of the organization was positively impacted by the
efforts ploughed into training and development by the organization
over the last few years.
(Source: T.K. Srinath, Executive Vice President & CHRO, Vistaar Financial Services Pvt.
Ltd., Bangalore.)
 Appendices 197

Appendix F

Why Perception is Important for People Analytics

In an article in The New York Times in May 2015, Alex

Peysakhovich and Seth Stephens-Davidowitz, data scientists with
experience at Google and Facebook, discuss “How not to drown
in numbers.”
In the article, they describe how big data will get us so far, but
it will never provide a full picture. With event data, we can under-
stand what has happened and predict what is likely to happen in the
future. With data about the individual for example demographic
data, we can identify who it is likely to happen to or who will
behave in this way in the future.
In some situations, just knowing what will happen and to
whom might be enough. In marketing, it might be good enough
to understand, “if I send this communication to these groups of
customers I am likely to see X return.” With employees this is
unlikely to be true.
Many of the models we develop for clients have an exploratory
data purpose to them. Clients are interested in knowing how the
model identifies individuals. What are the factors that will most
determine success or failure? If I do not give my sales person in
China a promotion within the first 22 months in a role, are they
likely to leave?
There are two great ways of dealing with this issue. Analysts
can use their domain knowledge, ideally based on the knowledge
of academic research, to understand why the effect that they are
seeing could occur. Alternatively—or even better, as well—they
can ask individuals why they have behaved as they have.
In The New York Times article, the authors describe how
Facebook and Google do this to understand user behavior. If a user
clicks the back button shortly after landing on a page in Facebook,
they might be delivered a short question on why the page wasn’t
of interest. As analysts, we need this why to move from an insight
to action.
198 Winning on HR Analytics

The Need for Perception Data

Collecting perception data is one form of data acquisition. Usually,
as people analysts we find that the best way of improving our initial
models is by including new data and not using better algorithms.
Data acquisition is a strategy which is not frequently discussed
in the popular discourse on people analytics, where people find
magical insights from the vast amount of existing data, but it is a
critical component for most experienced analysts.
Usually, these data need to be linked at the individual level to
other data we have from our HR and business systems. When this
is done, it can be used on everything from employee dashboards to
creating predictive models. It provides insight which we otherwise
would not have had.

Traditional Surveys Have Issues

HR has been capturing perception data for a long time. From early
job satisfaction surveys to more recent employee engagement and
from exit surveys to manager feedback it is rare to find a large HR
function which is not using surveys at least once a year.
Unfortunately, the value of these surveys is usually low.
Employees do not enjoy completing them. They are usually long,
comprising of tens of Likert-type questions covering every pos-
sible aspect that the designers could identify and often many more
that internal politics dictates should be included.
The results often are not terribly useful. Do we really care that
some measure has moved by 2% since last year? Is a 7% difference
when comparing two functions really important or is the result we
are seeing due to the fact that one department mostly has long-
tenured professionals whilst another has a young team, hired in the
last 24 months? Many of the statistical techniques used depend on
the assumptions that are frequently missing.
And whilst having 40, 50, or 60 questions might give an
illusion of detailed information the uncomfortable truth is that itis
usually less useful. In one experiment with the employee survey
data of a large firm we removed 80% of the answers. We then
 Appendices 199

used a recommendation algorithm to make predictions of what

the missing answers were. We got almost 90% right. Employees
answered in patterns.

The Joy of Text

One of the lessons that I learnt earlier was that executives love the
open text answers to surveys. We can weave a bunch of charts into
a narrative to show what is happening but one emotionally charged
comment can easily swing the decision.
The problem with text of course is that it does not scale well. It is
okay reading tens of comments but when you have tens of thousands
of employees globally, responding in multiple languages, making
sense of this data has been difficult or at least very expensive.
Fortunately, text analytics is now at a stage where it can be of
assistance. There are two key tasks that analysts typically want to
do with employees comments; we want to categorize them, that is,
identify what proportion are about each topic and in what context
they are being discussed, and we want to score them, for example,
rate how happy or sad they seem.
If we can accurately do this, we can transform the survey. A
typical engagement asks a small number of questions to determine
a level of engagement and a large number of questions through
which the analyst hopes to identify the factors that drive this
engagement.
Let us take the example of a recent set of survey data which we
looked at. Using an unsupervised learning approach, we were able
to identify 54 separate topics—both the subject such as “career
opportunities” and the context “shortage of career opportunities.”
Many of these entities would be expected by a good researcher
but others, for example “the parking situation”, would be highly
unusual to find on an employee survey.
This survey had asked two open text questions:

• If you could do one thing to improve your working life at

“company,” what would it be?
• What is the best thing about working for “company”?
200 Winning on HR Analytics

Both questions not only are asking for topics but also have a
ranking or importance aspect to them. This is difficult to get via a
traditional survey where there are technical issues to the ranking of
a large number of features.

Of course a largish proportion of individuals mention several

topics for each question. We, therefore, can look at co-occurrence
between topics—which topics are most likely to appear together.
Even only with single topic responses to each question, we are able
to identify how people answer the two questions in relation to each
other and other more traditional questions such as an engagement
score.
If we link the perception data to the demographic data or event
data, we are able to use algorithms to spot groups of employees
most likely to mention each topic. We can start to develop rich
stories about a multitude of employee perceptions.
At an aggregate, creating and analyzing the metadata associated
with the comments in this manner provide the ability to understand
trends and patterns individual text comments which frequently
communicate nuances that algorithms are unlikely to identify.
As Ben Shneiderman famously noted in his information seeking
mantra, “Overview first, zoom and filter, then details-on-demand,”
users need to have the ability to drill down to the base information,
in this case the actual comments. Good interactive data visualization
can provide this capability and the pattern-seeking algorithms can
be used to guide users’ attention to the important groups to make
the “zoom and filter” more effective.
There are several other places where people analysts have
access to large quantities of text data. Generically, the approaches
are similar—preprocess the data, score, or categorize it and then
use the metadata as you would otherwise, especially count data.
There are numerous packages and libraries available that make
many of the key tasks easier such as the tm package in R. There are
also several application program interfaces which provide various
aspects of the text workflow.
People analytics arguably suffers at the moment from inflated
and often unrealistic expectations. In many instances, this is
 Appendices 201

because the audience expects it to reveal more than just the

probabilistic patterns. With perception data we can move one step
further towards genuine insight, developing an understanding of
why our models might be behaving in the way that they do.
(Source: Andrew Marritt, Founder, Organization View GmbH, Zurich, Switzerland.)
202 Winning on HR Analytics

Appendix G
Case Study for Talent Acquisition
An Australian EPC company has revenues in excess of $5 billion
and employs more than 30,000 people. This company needed to
hire 150 plus specialist engineers to bolster their Middle Eastern/
North African/Indian engineering services capabilities. Niche skills
required include cross-country pipeline design, undersea cabling,
and offshore rig establishment. On releasing an advertisement, the
company received 19,200 applications. With their existing pro-
cesses and bandwidth, it would have taken them four months and
cost over a million dollars to successfully complete the process.
Using context-based search and analytics, they were able to accom-
plish the entire process in 45 days at 40% of the estimated cost.

Usage Context-based Search and Match

We search for something on the Internet all the time. The search
engine takes the phrases we have entered and then compares it
against the documents that have the same exact phrase. However,
the outcome we get is not sorted just on the basis of frequency of
the appearance of these keywords. The algorithm also searches
for documents that have been accessed by most people so that it
throws up the most relevant matches right upfront.
Resume screening, is a search activity as well. There are
keywords that a hiring manager is looking for. It could be,

• Years of experience,
• Skillset,
• Nature of projects worked on,
• Management experience,
• Location, and
• Quality of companies worked with.

Often the screening process takes time as the search is done man-
ually, resume by resume. To some extent, this can be accelerated
by searching for the keywords. However, the benefits are only
 Appendices 203

incremental as the screener still needs to open each resume and

then read through it.
Spire Innovations is a Bengaluru-based technology company
focusing on contextual search and intelligence. The company has
created the context for effective talent acquisition by creating a
context cloud for thousands of jobs. Once the hiring specifications
of the company are finalized, the company runs its context search
to arrive at a shortlist that is based on the demand for the jobs a
company is looking for. The effectiveness of this process can be
identified from the following visual.
In the first round of screening, 35% of applications were
matched. This is not different from a manual process but it took
just 48 hours. This was further pruned by 24% using the context
matching. The context was created by the company identifying
specific competitors within the industry.
Subsequently, the list was winnowed to just 370 most eligible
candidates, from whom 287 appeared for the interviews and 209
were made offers for employment. Using the technology, there was

Source: Spire Innovations Inc.

204 Winning on HR Analytics

more than 80% improvement in the effort needed for recruiter and
for panelists.
There are several steps within the talent acquisition process
which are ripe for the use of technology as this illustration shows.
However, where is analytics in this?
Spire Innovations have also created the dashboard for reviewing
the process to improve the talent acquisition process. This is
represented in the enclosed graphs.
Let us look at the first of the graphs. The first graph shows
the percentages of resumes from different companies that have
made the first cut and the second. For companies A, D, and F,
the gap between the two is less when compared to the other three.
This gives an indication of the suitability of profiles from such
companies that help in fine-tuning the effort.
The second graph compares the company’s compensation
with that being stated by candidates from different companies.
Other things being equal, this gives an independent validation of
competition salaries as well as the gap.
Third graph is around the notice periods stated for different
companies. Plotting and reviewing this helps a company to
consider this key input for selection. If one needs 10 people in a

Figure G.1 (a) Candidate Skill Matching

 Appendices 205

Figure G.1 (b) Competitive Compensation Intelligence

Figure G.1 (c) Notice Period Distribution of Interviewees

206 Winning on HR Analytics

Figure G.1 (d) Interview-to-select Conversions

month’s time, it makes little sense to make offers to employees

from companies that have a 3-month notice period.
The fourth graph presents the company-wise statistics on the
recruiting measures:

• By plotting the candidates who were called for interview

to those who appeared, one is able to get a measure of
comparative brand versus that company.
• By plotting the interview to offer rate, one gets an indication
of quality of talent available from that company.
• Finally, by plotting the offer to join rate, one gets the strength
of comparative positioning and likelihood of success from
hiring from different companies.

These measures are identified and sometimes tracked by

companies. However, they do not have the desired impact, unlike
this example, as the following steps are not taken:

1. The questions are looked at in terms of “how many” and

“how soon.” “Where from” is not looked at as an input for
 Appendices 207

strategizing the talent acquisition strategy. Often, recruiters

discern patterns, but lack the analytical language to articulate
the same.
2. The measures are stacked wrongly. In this example,
measures, such as cycle time, turnout rates, etc., are mapped
against unique positions against companies. This again helps
the process to be adaptive based on patterns identified. On the
other hand, often companies measure these across positions
or arrive at one particular number for the entire company. So,
a measure like “we have a 65% offer conversion rate” does
not help, when one is looking to hire for a specific position.
The numbers vary widely across the board and unless these
are tagged correctly, one might not be able to predict the
outcome of actions.
3. Looking at an available talent as a marketplace where the
company competes for resources using its value proposition
with other competitors. The ability to attract talent from
competitors is a very good indication of whether the
positioning is effective or not. A company-centric inside-
out view is important, but a market-centric outside-in view is
crucial for perspective as well as dynamism in HR policies.
(Source: Spire Technologies.)
208 Winning on HR Analytics

Appendix H
Case Study for Building a Business Case for
Employee Retention1
Infosys Limited is a global leader in consulting, technology,
outsourcing, and next generation services. The company has
revenues in excess of $9 billion and employs close to 200,000
people across the globe. The company has one of the largest
training facilities in the world in Mysore, India. To facilitate
ongoing growth requirements, the company hires thousands
of fresh engineers from across India and the world. The fresh
engineers undergo their three-month orientation to programming,
soft skills, and computer technology in Mysore.

Context
The company handles a challenging process of arriving at balancing
the following:

• Business needs for fresh engineers: This forecast is nearly

18 months ahead, but reviewed on a continuous basis, so
that most trainees can be allocated to projects on completing
their training.
• Skill requirements: The company trains fresh engineers in
several technology streams from Mainframe to. Net, Java,
etc. The allocation to skills has to be synchronized to the
skill requirements from the business.
• Allocation to locations: The software development headcount
in Infosys is distributed across 11 cities. The demand–supply
matrix has another variable in terms of utilization numbers
from each location.

1
 The data provided in this case study about Infosys is for data representation
purposes only. It is not to be republished in any form without the consent of
authorized persons from Infosys Limited.
 Appendices 209

The allocation of employees to the skill is based purely on business

demands. At the same time, the company used to consider the
employee’s choices of location before making the allocation.
In 2012, the company analyzed and found that the fresh engineer
attrition was having an impact on the business. The company
undertook a six sigma process to measure the reasons for this
attrition and reduce the same.

Root Cause Analysis

With available information, the company performed an analysis on
the causes for this attrition. The following reasons were identified.

1. Lack of alignment between the Infosys locations and hiring

locations. The company has larger centers in places, such
as Bengaluru, Hyderabad, and Chennai. The centers in
locations, such as Chandigarh, Gurgaon, etc., are relatively
smaller. The hiring numbers from different geographies
needed to integrate with the relative headcount ratios,
keeping diversity in mind.
2. Low awareness of demand-supply challenges. The trainees
were not really aware of how the demand for locations was
arrived at. Without this awareness, once allocated they
demanded transfer on the allocation being made.
3. Disconnect between actual demand and employee prefer-
ence. As per the existing process, the trainees needed to
make five choices from among the existing locations. This
gave them a feeling that they would be allocated to at least
one of these locations. In reality, the demand may be in five
other locations only. This made the trainees disillusioned
with the process.
4. Inadequate communications on the allocation process.
The trainees are fresh off college. Just doing a single
communication and mailers to them were insufficient, as
they were also transitioning to the processes of a corporate
career for the first time.
210 Winning on HR Analytics

Objectives
The following are the actual measures and goals set by the
company.

S. no Measure Actual Objective

1 Percentage of top three preferences 73 82

met

2 Percentage of top preference met 47 57

3 Employee satisfaction with the 44 60

process

Actions Taken
In line with the root cause analysis, the following actions were
taken:

1. Repeated communication on the allocation process. The

trainees were given an initial overview as well as a mid-
course refresher on the allocation process. FAQs and SOPs
were shared on portals.
2. Transparency into the list of available vacancies. The number
of vacancies available in each location was shared with the
employees. This increased their trust in the process, as they
were no longer giving preferences based on just their wishes.
3. Simplification of the preferences. Instead of five, trainees
needed to just identify their top three choices. The process
was initiated on the basis of demands 1 month before the
final allocation. This was revisited 15 days later, with the
most current set of demands. Trainees were allowed to
change their choices accordingly.
4. Creation of an algorithm to prioritize allocations. It is possible
that 100 trainees may want to be in Bengaluru, while only 60
slots were available. An automated spreadsheet was created
to prioritize in such cases based on the first preference as
well as the performance during training. In other words, a
 Appendices 211

superior performance during training is the best enabler of

getting a location of choice.
5. Involve all stakeholders on a continuous basis. Regular
updates were shared with the location and BU HR partners.
The numbers were shared and process was reviewed with
all stakeholders, namely corporate planning, talent planning
managers, HR, Global Education Center allocation team,
etc. These meetings helped surface the improvement
opportunities.

Results
In the quarter before the new process was rolled out, the attrition
was 2.74%. The next batch was 4500 strong.

Batch Size: 4,516

Expected attrition at 2.74% = 124
Actual attrition: 59
Attrition percentage: 1.3%

The company also performed creditably against the objectives it

had set.

S.no Measure Target Achievement

1 Percentage of top three preferences 82 80

met

2 Percentage of top preference met 57 63

3 Employee satisfaction with the 60 58

process

The pilot showed a positive outcome. Subsequently, the changes

were institutionalized.

Business Benefits

What were the business benefits deriving from this exercise?

212 Winning on HR Analytics

Training is expensive, as there are not only payroll costs incurred

but also costs incurred in terms of infrastructure, training costs,
transportation costs, etc. These worked out to around $5,000 per
person. When a trained employee leaves, this becomes a sunk cost.
The new process helped reduce allocation-related attrition by
65 (124–59).
Straight off, this was a saving of 65 × 5,000 = $325,000.
However, there are more benefits as well, from reducing
attrition. If a trainee stays on with the company, there is a 70%
chance of him/her getting absorbed into a billable project. The
company potentially loses that revenue.
It is assumed that it would take six months to find a replacement
when a trainee leaves. The opportunity cost of reduced attrition
then becomes one of the revenues enabled. In this process, how
much did the company benefit from retention?
Taking a conservative estimate, let us take that they are billed
$3,000 per month. Then the billing loss if 65 more people quit
would be

65 × 0.7 × 3,000 × 6 = $819,000.

The business value from the initiative then is of two parts

1. Savings on investment: $325,000

2. Opportunity retained: $819,000

Summary

This concise case study illustrates the power of reviewing metrics,

setting numerical goals for a program and then establishing tangi-
ble RoI. Many HR processes can benefit from such an approach.
(Source: Infosys)
 Appendices 213

Appendix I
Using Statistics to Arrive at Engagement Drivers1
App Annie is a company focused on delivering powerful market
data and insights that allow companies to succeed in the app
economy. They are no stranger to how analytics can help people
and companies make better decisions. Their free business analytics
dashboard aggregates app store, advertising, and usage data for
a single view into an app or app portfolio’s key metrics. Their
app market data provides powerful insight into the market and
competition including store intelligence to view download and
revenue data, usage intelligence to understand an app’s usage,
reach, and retention, and audience intelligence to view user
demographics and cross-app adoption.
Founded in 2008 in Beijing, they are now headquartered in San
Francisco with 400 plus employees across 15 global offices (in four
different continents). Denise Lyle is the Global People Programs
Senior Manager at App Annie (part of the Human Resources
team), responsible for managing the HR related programs that help
App Annie grow and scale. Their strategy in building out these
programs involves the use of people analytics on the Culture Amp
platform. Lyle says that the need for analytics is great as they are
a high-growth global company. In her words, “We are definitely
dealing with a lot of moving parts at the same time in different
corners of the world.”
As a global organization, App Annie’s values reflect the spirit
of the company and its employees, and it strives to live up to them
in everything it does. Whether it is “making it happen” on a tight
deadline, having a festive happy hour after a product launch to
“empower individuals and celebrate achievements,” or working
across its global offices with the talented people that make up its
“many cultures, one team,” App Annie’s values can be found at the
root of all of its decisions.
When App Annie employees were asked which value they
personally identified with. Lyle says that about half of the

1
 Learn more at www.appannie.com
214 Winning on HR Analytics

company said “Many cultures, one team” which she attributes to

the fact that, “we certainly do have many cultures here, but there
is this sense of ‘We are all in it together.’ We all share something
culturally even though we may have this ethnic or geographical
difference.”

The Challenge
App Annie’s major challenge was that growing fast made it difficult
to drive the right behaviors to achieve the expected outcomes,
and their internal and external brand was suffering, explains John
Chu, Vice President of People at App Annie. Also like many large
companies, especially those with a distributed team, they needed to
ensure their employees continued to be engaged; they have grown
from 45 employees in 2012 to over 450 employees in 2016. App
Annie’s first iteration of trying to understand employee sentiment
was a brief engagement survey and series of culture focus groups,
led by local HR leaders, across the company. These were small
group discussions on the culture of the company with key focus on
identifying App Annie’s core values. While this provided the HR
team with some data, they wanted to dig further to find out more
about engagement in their employee base.
During this time, App Annie received some negative reviews on
the recruiting website Glassdoor. This issue was later determined
to be caused by a fragmented culture and shift in expectations as
the App Annie headquarters was moved from San Francisco to
Beijing. Lyle says quite frankly,
[W]e don’t want negative reviews on Glassdoor for two reasons. Primarily,
we don’t want people to feel that way, we don’t want them to have a poor
employee experience at App Annie. Secondarily it also harms recruiting
efforts and our employer brand. If we’re not providing employees with a
way to give us feedback directly, other than this extremely public forum,
something’s wrong.

Looking back on this time Lyle says, “We just wanted to know
what was going on. We thought we knew, but the outputs were
telling us some of what we knew but not everything. We needed to
know what was happening with our people.”
 Appendices 215

The Solution

App Annie decided to use Culture Amp, the world’s first real-
time People Analytics platform, to collect and analyze data
from engagement surveys. In her own words, Lyle says, “We
use Culture Amp to help us figure out what’s going on, gauge
employee sentiment, see what’s working, what’s not. Because we
are growing so fast it’s really important that we’re agile and we
iterate quickly.” One instance in which the ability to iterate quickly
shows itself is in App Annie’s practice of setting quarterly goals to
keep them moving at a pace to meet their annual goals. This allows
them to shift priorities and “focus on what matters most or what
needs most attention using Culture Amp analytics,” says Lyle.
App Annie’s survey methodology is to take two engagement
surveys on an annual basis. One main, 50 question survey in Q4
followed by a 15–25 question “pulse survey” in Q2. Their end-of-
the-year survey acts as their baseline with a wide range of questions
to inform key areas for planning and action while the pulse survey
gives feedback on the actions that have been taken, provides index
information, and potentially identifies timely issues. High-level
results are shared at an all-hands meeting one month after the
survey closes and planned action is shared the following month
after multiple executive team strategy meetings. Onboarding and
Exit surveys are implemented on an ongoing basis.
Where Culture Amp really helps the HR team is in revealing the
drivers of engagement, which help interpret the results into tailored
actions. The statistics behind the driver analysis helps the HR team
at App Annie accelerate the typically lengthy duration of analysis
to understand what to address and take action faster. Drivers of
engagement are a Kendall tau-c correlation analysis that identifies
statistical dependence in ordinal data. Rather than focusing simply
on high or low scores, it relies on correlative statistics to identify
which topics would have the largest impact on engagement and
business outcomes. Lyle says specifically, drivers “provide rich
insight for us because they show us what matters most to our
employees and where we should focus our efforts to make the
biggest impact on engagement.”
216 Winning on HR Analytics

Culture Amp helped App Annie learn their overall drivers of

engagement and allowed them to drill down to drivers in each
region:

This is insightful for us because what is working well in some regions may
not be working as well in others, and we can target some of our action plans
geographically. For example, we had a surprising result that our employees
in APAC were more engaged with our leadership, like communicating
the vision of the company, than our employees in the headquarters in San
Francisco. That was an interesting data point because you would think that
where leaders are located, there would be stronger engagement but we
actually saw the opposite. This showed us where to dig a little bit further to
understand some root causes.

Additionally, survey items in the leadership factor show a high

Kendall tau-c coefficient, ranging from 0.34 to 0.52; signaling
these topics relate to business outcomes like productivity and as
App Annie has evolved their people analytics methodology Lyle’s
attitude towards surveys changed as well. She says,

I think what’s most important with gathering survey data, and Steven
Huang (Strategist, Data and Insights at Culture Amp) taught me this, is
data, insights, and analysis are helpful. But what’s really important is how
you use the data to plan and take action. It’s vital to communicate with your
employees along the way.

After their most recent annual engagement survey, App Annie

held a global meeting with their senior management group to
discuss the results. Each leader received access to specific parts
of the survey (divided by three regions: Europe, America, APAC
and their department) to review beforehand. They were asked to
arrive prepared to discuss some of their hypotheses on engagement
based on the survey data, along with a plan of action. Lyle says,
“We spent a few hours in breakout sessions facilitated by our HR
leaders, looking at specific data and then coming up with action
plans by department and by region. Once action plans were created,
we communicated back to the employees.”
 Appendices 217

Business Result

Chu says one of the biggest benefits of using Culture Amp is that
the results have helped to “inform and validate leadership decisions
to change multiple organization structures and move forward with
a larger investment in strengthening manager effectiveness”. It also
allowed the HR team to get executive buy-in for people programs
at App Annie. As Lyle says,

Having data that is valuable to the executive team is really key in making
any sort of progress with action planning. Otherwise it is just an HR
initiative. It’s a very tight window of opportunity right out of the gate to
demonstrate data that has strong insights, that is valuable, that is easy for
an executive to understand and digest.

A top driver of engagement at App Annie is the leader

communicating a motivating vision about the company (Kendall
tau-c coefficient of 0.52). Based on their most recent survey
results, Lyle says they are on the lower side, but there is action
planning taking place to bring those scores up. App Annie plans
on sharing their company vision more openly, more frequently,
and getting it in front of the employees at company meetings.
Specifically, this feedback spurred the creation of their two-part
annual survey methodology and skips level one-on-ones. So, in
addition to executives and managers having one-on-ones with their
direct reports, they are encouraged to have conversations with the
employees of their direct reports.
Skip-level meetings can be the best way to learn about your
team and a wise investment of your management time for these
reasons. First, leaders can gain insight into what is going on with
their teams and their day-to-day work. Second, they can assess
how managers are doing by checking in with their employees.
Lyle says,

[I]t’s also an opportunity for leaders to provide transparency—hitting

directly on that issue of communicating a vision that’s motivating. It
enables leaders to get real-time feedback about what the company is
doing, what we should and shouldn’t be doing, and getting the employees
perspective.
218 Winning on HR Analytics

They are also encouraging check-in meetings with the HR team

and traditional manager one-on-ones.
Using Culture Amp has allowed App Annie to collect valuable
data on their people and allows them to make better decisions to
scale their company. Lyle says,

As the People Operations team in a fast growing organization, we have

to use the analytics in order to plan, iterate, grow. Insights and analytics
are key to the growth of our company and making sure we are moving in
the right direction. We certainly know we are moving fast, we just need to
make sure we are focusing our efforts in the right place—we use analytics
to do that.

(Source: Steven Huang, Strategist, Data & Insights, Culture Amp, San Francisco,
California, US.
Alexis Croswell, Associate Marketing Manager, Culture Amp, San Francisco, California,
US.)
 Appendices 219

Appendix J

Making the Case for Predictive Attrition Risk Modeling:

A Roadmap for the Future

G.D. Graham, E. Olesen, and R. Dutta

As organizations continue to move up the “people analytics
maturity curve,” the capability and appeal of using advanced
statistical modeling techniques to predict who is likely to leave
and why is becoming more prevalent. However, creating statistical
models to predict attrition risk is not enough. Two critical issues
prevent organizations from harnessing maximum value from these
efforts. First, you cannot sacrifice on robustness of your models.
This may seem obvious but often organizations overestimate the
validity of their data, and in doing so inevitably compromise the
integrity and utility of their predictive models. When adopting
predictive analytics, many organizations make the mistake of
oversimplifying the process by running bivariate correlations
on a handful of HRIS data fields, often focusing on a single data
source, or performing simple psychometric assessments. Unless
the organization adopts a more “complete” stance by gathering
data from different sources, internal or external, and then validate
their model predictions against actual results, credibility, and value
delivered by these models will remain questionable.
The second issue is equally important and perhaps even more
challenging: organizations must actually do something with the
results from these models. They need to find a way to integrate
results from their predictive modeling efforts into management of
talent workflow and start to address the “so what?” question that
arises once the models are built. Actions derived from predictive
models can have impact on change management, hiring strategies,
workforce planning, risk management, and IT integration. This
brings a wealth of new challenges, but neglecting to act on results
eliminates any possibility of getting real return on investment of
predictive modeling.
Although more and more organizations are investing in
predictive attrition risk modeling, very few organizations have
220 Winning on HR Analytics

totally and seamlessly integrated their results into their talent and
workforce management practices. That said many organizations
are at least starting that journey—whether it is improving data
governance to use for a predictive effort, or conducing pilot studies
on targeted populations. Other organizations, like the case study
presented in this chapter, are further along on the journey and
have validated attrition risk models over several years and begun
action planning based on results. The following chapter provides
guidance around optimizing a predictive analytics program and a
real-life example of an organization currently capitalizing on the
potential of predictive attrition risk modeling.

Step 1. Ensuring the Model(s) Is/Are Robust

When embarking on the predictive analytics journey, it is important

to remember the tenet of “garbage in, garbage out.” In other words,
while logical processes may be robust, the outputs of predictive
models are only as good as their data inputs and/or methodological
approach. Therefore, any actions or initiatives that stem from those
models will only be effective if the models themselves are reliable
and that includes not just statistical validity measures like area under
the curve or Kolmogorov–Smirnov(KS)value but actual behavior
of employees compared against predictions. If organizations truly
want to use predictive modeling to affect change, then the model(s)
must be rooted in sound data, techniques, and methods. Ensuring
model robustness is also key to building credibility in the model
and confidence from business leaders, both of which will be critical
in building the business case for change.
Start small. Rather than creating an attrition risk model for
the entire organization, it is advisable to focus initial efforts on
a specific pilot population, such as a business unit, job role, or
location where attrition is an issue. While the selected group will
still need to be large enough to generate reliable results, focusing
on a targeted population is not only more manageable from a data
governance perspective but can also give more precise insights into
the attrition risk factors affecting that particular group. Even once
the pilot population has been selected, HR leaders should also think
 Appendices 221

critically about other criteria for inclusion. For example, a client-

facing organization might want to exclude internal support teams.
Additionally, part-time employees or those that have been with the
organization for less than a year typically leave for different reasons,
and therefore may also need to be excluded from the pilot study.
Even within a pilot sample, certain subgroups may act differently
when it comes to attrition. For example, within a key role, junior
staff may attrite at a different rate than more senior staff. In some
cases, these differences are great enough to warrant separate
models. Prior to modeling, practitioners should statistically
investigate differences between groups in attrition risk variance
to determine if multiple models should be deployed to explain
different segments of the pilot population.
Only use reliable data. In gathering data for predictive modeling,
it can be tempting to use any information or metric that is readily
available. However, data that is not systematically and reliably
collected can result in erroneous findings. Data that is either not
cleanly available or not reliable should be omitted but also tabled
as a data governance objective for the future.
Collate multiple sources. Data is also more valuable when it
is gathered from different sources across the organization. Using
a single source of data to test your hypotheses not only limits the
statistical viability of your dataset but also provides a narrow
set of predictors for testing your hypothesis. In addition to the
standard information found on a typical HRIS platform, analysts
should think creatively about other sources that can be leveraged
from both inside and outside the organization and those decisions
should ideally be guided by a hypotheses framework, rather than
arbitrary data hunting. Within the organization, access other
databases such as learning management, performance ratings,
and employee surveys. Outside the organization, macroeconomic,
socioeconomic, social media, and benchmarking data can also
supplement your list of predictive input data.
Look beyond bivariate relationships. The end goal of an attrition
risk model is not to identify any and all factors that correlate with
turnover. Simply reporting the factors that correlate with attrition
will result in a list of correlates that may not have any causal linkage
222 Winning on HR Analytics

with attrition behavior. Instead, attrition risk modeling should use

a combination of statistical criteria and organizational knowledge
to expose the predictors of attrition that are most impactful. This
not only makes findings and results more manageable in terms of
reporting but also provides insight into the specific levers to pull in
order to affect changes in turnover rates.
Verify against actual attrition. Rather than immediately acting
on pilot model results, organizations can do much to ensure the
validity of their model by letting it “play out” for a while and then
compare the model-generated attrition predictions against actual
attrition behavior. Just as inaction can detract from the utility
of predictive analytics, premature action implemented before a
final model is solidified can be equally harmful. Therefore after
conducting a pilot study, many organizations spend six months
to a year simply observing the model and then determining the
percentage of voluntary leavers that were correctly identified as
“high risk.”
Refine as needed. As the organization continues to change,
the model should continually be refined to remain valid and
useful. This could mean that new data has become available that
should be tested, or existing data should be re-operationalized in
a more meaningful way. Additionally organizational initiatives,
population turnover, and even socioeconomic factors could render
certain predictors more or less impactful in characterizing attrition.
Therefore models should be continuously evaluated in terms of
validity and refined as necessary.

Step 2. Integrating Model Results into Organizational

Talent Workflow

Once model results have been validated and refined, HR can

begin to take actions to mitigate attrition. First, findings from
attrition risk modeling should be integrated into the organization’s
retention strategy wherever possible. This means identifying the
actionable factors in the model that are causing employees to
stay, and capitalizing on them in various aspects of the employee
experience. For example, if training hours is associated with lower
 Appendices 223

attrition, HR’s learning and development sector can find ways

to make training more accessible, useful, and prolific among
employees. In the same vein, HR can identify actionable factors
that increase attrition risk, and implement ways to mitigate them.
If career stagnation has been shown to increase attrition initiatives
such as cross-job training, coaching, and reward structures can all
be avenues for improving retention.
In terms of talent management, results from an attrition risk
model can be a key piece of information for leaders. Using robust
statistical algorithms, attrition risk modeling can predict the
likelihood of attrition for all employees in the organization, and
classify “high risk” leavers as necessary. This list of probable
leavers can be mapped against the list of top talent that leaders
generate during an annual talent assessment. Taking the outputs
of attrition risk modeling one step further to identify not just the
potential loss but potential regrettable loss can help leaders plan
their talent management strategies accordingly.
Attrition risk findings can also inform selection strategies.
While some predictors of attrition, such as educational background
or previous experience, may not be actionable in terms of the
current employee population, they can be targeted in recruiting,
assessment, and selection. In doing so, organizations can weed out
candidates with high-risk characteristics and focus on those with
characteristics consistent with longevity in the organization.
Finally, HR, and IT sectors will need to work together to answer
the question: How can we integrate attrition risk findings into
our technology infrastructure? To begin, dashboards and other
visualization tools are useful for monitoring significant predictors
and attrition patterns. Many organizations start this effort by
creating a stand-alone attrition risk dashboard, and work towards
incorporating it into the overall HR dashboard. IT can also enhance
data governance structures to create easy access to modeling
data. Currently, a significant amount of manual work goes into
pulling, merging, compiling, and computing modeling data, but
in the future the more sophisticated organizations will be able to
do all of this at the click of a button. Few, if any, organizations
currently have this capability and current “black box” approaches
224 Winning on HR Analytics

sold by some software vendors are not robust enough. But along
the predictive analytics journey the demand will ideally be for a
system (rather than a team of individuals) that can run models on
an ongoing basis and continually provide results to managers and
leaders of an organization.

Case Study: Booz Allen Hamilton1

Over the past 5 years, Booz Allen Hamilton (BAH), a large
professional services firm based in the US, has worked with
PricewaterhouseCoopers (PwC) People Analytics (Saratoga)
practice to build, refine, and act on attrition risk models as part of
their larger HR strategy. Within the company, attrition rates had
been high for some time but leaders had few tangible results on
which to base their retention strategies. As a professional services
firm, BAH recognized that people were their main “product,”
so any investment made to mitigate unwanted attrition is likely
to have a substantial return. Additionally, the company had a
reputation for being tech-savvy in many of their own offerings,
and wanted to apply that same scientific rigor to solving their
“people problem.” Predictive analytics—particularly, attrition
risk modeling—was identified as something that would be both
useful and a differentiator for BAH in the talent management
marketplace.

The Roadmap to Date: Years 1–3

In the first year of their attrition risk modeling effort, BAH

identified the need to create three models—one for each of their
three business units. This decision was based not only on the
statistical finding that attrition rates differed between the units but

1
 Booz Allen Hamilton provides management and technology consulting and
engineering services to a range of public sector organizations, private corpora-
tions, government agencies, and not-for-profit companies. With offices and
clients around the world, BAH provides expertise and insight in consulting,
analytics, technology, cybersecurity, engineering, innovation, and more.
 Appendices 225

also the practical realization that any actions or efforts stemming

from the results would most likely be implemented at the business
unit level. After the models were built, BAH focused primarily on
disseminating results to senior leaders and educating them on the
potential utility of the findings. This focus on gaining leader buy-in
not only prevented HR from implementing actions premature but
also helped strengthen the business case and support for attrition
risk modeling across senior leadership. The rollout of findings
largely consisted of education on the purposes of the model and
discussions around particular variables. Specifically, HR leaders
continually expressed that the model was intended to identify the
factors that affected attrition above and beyond all else, rather than
simply identify any and all contributors to turnover. Discussions
also centered on concerns from leaders, such as ensuring that
results were only distributed to only the high-level managers and
protecting the privacy of employees identified as “high risk.”
Beyond this education of high-level leaders, no other tangible
actions were taken other than simply observing the model over the
next year.
The second year brought refinements to the model as well as an
increased interest from leadership and a more useful communication
of results. Due to HR’s large-scale educational efforts in the prior
year, leaders who saw results last year began to ask for updated
findings, while leaders who did not see the results asked to be
included in the current year’s rollout. This brought the challenge
of educating a new segment of leaders and opening up the model
to more critique, but it also meant the demand for attrition risk
modeling was growing within the organization especially because
the predictions made in the previous iteration were increasingly
coming out to be true. In assessing the accuracy of the first year
prediction, BAH saw that most employees who had been classified
as “high risk” actually left the organization in the following year,
while very few “low risk” employees left. To accommodate the
demand and make results more useful, the second year results
included a rank order of elevated risk employees based on their
model-generated probability score. This helped leaders prioritize
who they wanted to target in their retention strategies. HR leaders
also created a toolkit to assist leaders in using the information.
226 Winning on HR Analytics

The toolkit contained the findings themselves, information on

how to interpret the findings, and a preliminary “action guide” that
leveraged information found in current best practice periodicals.
It also contained guidance and questions for managers to use in
conducting “stay interviews.” Managers were encouraged to
integrate these questions into their already existing one-on-one
cadence with direct reports, with a focus on those employees
identified as “high risk.”
BAH has recently completed its third annual iteration of attrition
risk model. This past year, BAH implemented changes to the model
itself, the communication of findings, and their strategic retention
actions. To start, the model was expanded to a larger population
within the organization. This decision was based not only in
confidence of the existing model’s robustness but also in demand
from other leaders whose business units had not been included in
prior models. The request signified a shift from “push” to “pull”
in terms of how information was distributed to the organization’s
decision-makers. Leaders were now directly asking for annual
attrition modeling results. HR and executives were continually
having discussions around attrition, and in turn were seeking out
as much information as possible. Attrition risk modeling results
had come to be viewed as a critical piece of that information.
To further aid in the dissemination of information, HR partnered
with IT to create a Tableau-based dashboard visualization of
attrition risk results. In prior years, results were given in static
Microsoft Excel files containing a list of high-risk employees
and heat maps that identified pockets of high risk across the
organization. With the deployment of an attrition risk dashboard,
the findings could be disseminated more efficiently and included
many of the details that leaders had previously been asking for,
including segmentation and filtering capabilities. Other HRIS
information was also integrated into the dashboard that provided
more information about high-risk employees which would be a
“regrettable loss” (e.g., top talent, high tenure or time in level,
recently promoted, highly billable, etc.). This information further
helped leaders prioritize who they needed to address in their
retention strategies. At the completion of the third year, BAH
maintains that it is still too early for enterprise-level action based
 Appendices 227

on modeling results. However, they noted that there has been a

tangible shift in focus when it comes to acting on attrition risk
models in which BAH leaders went from addressing teams of high-
risk employees to addressing all teams in the organization and the
high-risk employees within those teams.

Roadmap for the Future

As the results of the attrition risk models continue to effectively

inform retention strategies BAH plans to implement other predictive
studies that address a variety of organizational issues. First, with the
support of PwC People Analytics, they are preparing to conduct an
analysis to predict the quality of hire among current and potential
employees. The first step in that process will be to identify a
primary key role that will be critical for growth in the future. For
example, the role of software engineer is seen as a key position
and consists of about 1,600 employees across the organization.
By targeting a particular role rather than a family of jobs, BAH is
hoping to generate a specific quality of hire profile rather than a
more generic characterization of a “good” employee. The end goal
of a quality of hire model would be to identify the characteristics
that make someone successful at BAH and ultimately translate that
into a hiring “checklist” that can be used when evaluating potential
candidates. However, this model brings its own set of challenges.
For one, the outcome variable—quality of hire—is not as easily
defined as attrition. Leaders and HR practitioners would need to
think critically about what constitutes a “high-quality hire,” and
how those characteristics can be measured and operationalized.
Additionally, much like leader education was the initial focus of
the attrition risk model, significant efforts would need to be made
to orient leaders to a qualitatively different predictive model.
Fortunately, the progress made in educating leaders on the utility
of predictive attrition modeling will provide a solid foundation for
easily and efficiently getting leaders up to speed.
Another potential avenue for predictive analytics is a study
around why candidates reject hire offers from BAH. The interest
in this model was sparked by an uptick in this metric, but again
provides a unique set of challenges for HR to address. To date,
228 Winning on HR Analytics

the most critical challenges are surrounding both the quantity and
quality of existing data. Not only is there a paucity of data being
collected in this area by BAH but the data that does exist is largely
derived from self-report and is therefore less reliable than more
objective metrics. Nevertheless, as BAH continues to expand and
integrate its predictive analytics capabilities, leaders will be more
willing and able to take on such challenges in an effort to create
data-based actions and strategies.
Full integration of multiple predictive tools is also a long-term
goal for BAH. By marrying results from models around attrition,
quality of hire, declined offers, and more predictive analytics can
inform the entire workforce planning process. Retention strategies,
selection processes, career development practices, visualization
tools, and other aspects of talent management can all be rooted
in a series of robust statistical findings and data-driven decision-
making. This will provide BAH, and other organizations willing to
embark on this journey, with a more concrete foundation on which
to build their leading practices and a greater return on investment
from their workforce and workforce planning initiatives.
Finally, the ultimate goal of any analytics investment is to
eventually integrate analytics into all business decisions. Right
now, BAH, like many other organizations, uses data in most if not
all business decisions. For example, basic decision-making around
recruitment and selection needs can be informed by simple data
such as the number of new hires in last month. But to make more
long-term strategic plans and decisions, organizations will need
more than just straightforward data—they will need the insights
and knowledge that can only be generated by more complex
analytics. Right now, BAH is applying these predictive methods
to the beginning and end of the employee life cycle with the quality
of hire and attrition risk modeling, respectively. But what about
all of the aspects of the employee life cycle that fall in between?
Just as there is currently no part of the life cycle that does not have
data associated with it, a robust analytics program will eventually
ensure that predictive modeling and analytics touch each phase
of the employee lifecycle as well. In doing so, a comprehensive
analytics program can elevate data-driven business decisions into
robust strategic initiatives rooted in rigorous analytical insights.
 Appendices 229

Gia Graham, Ph.D., PwC People Analytics

Gia Graham is a Senior Associate with PwC in the people analytics

practice. Her work consists of various analytic initiatives such
as metrics and benchmarking, conjoint analysis, workforce and
customer surveys, and predictive modeling. Prior to joining PwC,
Gia worked at the US Army Research Institute for the Behavioral
and Social Sciences as a research fellow and research scientist.
Gia earned her PhD in Industrial/Organizational Psychology from
George Mason University in 2013.

Eric Olesen, Booz Allen Hamilton

Eric s a Senior Associate with Booz Allen Hamilton and leads their
Workforce Analytics function. He has over 15 years experience
in the fields of Applied Statistics and Predictive Analytics in the
areas of Human Capital, Quantitative Market Research, and Survey
Methodology. Eric received his bachelor degree in Psychology and
Business Administration from Wayne State University and holds
a Master of Science in Industrial/Organizational Psychology from
the University of Akron.

Ranjan Dutta, PhD, PwC People Analytics

Ranjan is a Director at PwC’s LLP with over 15 years of expe-

rience in management consulting, workforce analytics, statisti-
cal modeling, and behavioral economics. He leads the predictive
analytics practice at PwC People Analytics (Saratoga). Ranjan
has provided advisory services to senior executives all over the
world spanning multiple industry sectors and functional domains.
He is a thought leader in the area of people analytics and has been
variously interviewed/ quoted by prominent media outlets, such as
CFO.com, HR Magazine, HBR Analytic Services, HR Executive,
CNN/Money, Wall Street Journal, Bloomberg BNA.
230 Winning on HR Analytics

Appendix K

DISCOVERING TEAM COHESIVENESS AND INFLUENCERS USING

ORGANIZATION NETWORK ANALYSIS

One of India’s leading financial services companies was seized

of a challenge. Should it continue to fill its ranks by hiring from
outside for middle management roles? Or should it groom talent
from within, by hiring at entry levels? This is a problem faced
universally by organizations.
Growing from within creates a strong culture, but needs patience
and a mature approach to deliver results. Employees need to grow
into roles and get promoted. Need-based external hiring is swift at
the outset, but in the long run it takes far more effort to onboard
and integrates employees from different organizations. Long-term
retention becomes another challenge.
While the problem is common, the organization innovated by
using the emerging technique of organization network analysis.

Principles

Organizations are not what their formal structure denotes. Let us

look at a social network metaphor. In social networks, the stories
can emanate from anywhere and gain likes. Influencers in social
networks are not necessarily the biggest stars but ones with a very
original point of view. It helps to be a celebrity.
Similarly, work and goals cascade down a formal organization
structure. Increasingly, organizations resemble a network and the
successful ones are far more cohesive than those which are not.
The informal and uncharted organization wields as much influence
or more, than the formal.
Mapping information flow helps us in recreating the informal
organization. Correlating the cohesion of the informal organization
with the percentage of in-house/lateral talent would establish the
case to groom from within or otherwise.
 Appendices 231

Problem Statement
The company had moved from a “buy just in time talent” to “make
talent in-house.” However, the success rate for this strategy varied
across different business entities. Some of them attributed a talent
shortage to account for their region’s mediocre performance. The
company wanted to:

• Identify what the enabling conditions for the success of the

strategy are.
• Identify who can be tapped to ensure that the program is a
success.
• Assess the long-term viability of the strategy.

The company worked with i-Cube, a company specializing in

organization network analysis using their proprietary tool OWEN.

Approach
The company ran a survey across its HR organization. The brief
survey required the participants to answer questions across areas,
such as mentorship, innovation, career, etc. In the survey, there
were specific questions for employees to enlist who in their peer
group is:

• Someone who they interact with for helping out with their
daily issues,
• Someone who they look upto as mentors, and
• Someone who is their role model and whose inputs they
value for career growth.

That someone can be their manager but also a peer or a manager in

a different group. Based on their inputs, the informal organization
was created. The following insights emerged.
232 Winning on HR Analytics

1. Uniformly distributed networks perform better

Let us look at the two entities. Each point corresponds to an

individual and the connections are based on the survey inputs. The
one on the left shows a more fluid structure with a greater degree
of interconnects. On the other hand, the structure on the right has
fewer nodes through which the information flows and there are
outliers, who are not connected to anyone. This correlated with the
performance of respective organizations.
It was also found that the closeness of a leader to his/her team (as
measured by the average distance in nodes for all team members
from the leader) also influenced the performance positively.

2. Cohesive teams perform better

Cohesiveness of the team can be represented using the following
two measures:

1. Team density: Let us assume that there is a 20-member team.

Let us consider two scenarios. One, where only 5 members
are identified as “go-to” people by the team. The second,
where 15 out of 20 people are mutually identified as “go-to”
people. Team density in the first case is 5/20 = 0.25, while
in the second case it is 15/20 = 0.75. A higher team density
indicates greater collaboration levels within the team.
 Appendices 233

2. Team average path: To begin with, each team member is

just a point on a plot. Then, based on the survey feedback,
connections are established between points. Some points
have several lines passing through them, indicating an
influential role. How many degrees of separation are there
on average for every team member? A team with an average
path of 1.2 is far more cohesive, than the one with 2.1 for
instance.

The organization’s survey showed the following trends. These

clearly established that there is a correlation between cohesiveness
and performance.
234 Winning on HR Analytics

3. Junior team members can wield disproportionate influence

Once the linkages are plotted, we can identify the influencers by

their connectedness. “Connectedness” is indicated by the number
of times a certain node (person) appears in the shortest path
between two other nodes. A well connected person will have a
higher score of connectedness, than a lower one.
The study identified several team members, who wielded
considerable influence.

Summary

By using the organizational network analysis, the company was

able to identify the factors that can make or break the HR programs,
independent of the merits of the program. In a loose network, the
leader’s formal power and his/her opinion holds sway. If the person
 Appendices 235

is in favor of lateral hiring, then lateral hiring works. However, in

a more cohesive team with distributed influence, there are greater
interactions and the team arrives at the best decision based on the
team’s view. HR can leverage this by sharing the business case
not just with the formal organization, but by reaching out to the
influencers also.
Network analysis offers us a powerful way of measuring and
visualizing the strength of teams. With future going towards
holacracy from bureaucracy such analytical tools are expected to
play a big role.
This case study is on the basis of inputs provided by M/s I-cube,
using their OWEN framework.
(Source: I-cube)
 References

Adler, Seymour, & Golan, Jacob. (1981). Lateness as a withdrawal behavior.

Journal of Applied Psychology (October), 66(5), 544–554.
Baird, L. & Meshoulam, I. (1988). Managing two fits of strategic human resource
management. Academy of Management Review, 13, 116–128.
Barney, J. (1991). Firm resources and competitive advantage. Journal of
Management, 17(1), 99–120.
Bersin, J. (2016). Is software better at managing people than you are? Fortune.
Bersin, J., O’Leonard, K., & Wang-Audia, W. (2013). High impact talent
analytics: Building a world class analytics and measurement function.
Oakland, CA: Bersin by Deloitte. Retrieved July 18, 2016, from http://
marketing.bersin.com/rs/bersin/images/hita100113sg.pdf
Cantrell, S., Benton, J.M., Laudal, T., & Thomas, R.J. (2006). Measuring the
value of human capital investments: The SAP case. Strategy & Leadership,
34(2), 50.
Cascio, Wayne F. & Boudreau, John. (2011). Investing in people: Financial
impact of human resource initiatives. New Jersey: Pearson Education.
CedarCrestone. (2014). Going global with HR technologies—2014:
CedarCrestone 2013–2014 HR systems survey. Alpharetta, GA: Sierra–
Cedar.
Coco, Cedric T., Jamison, Fiona, & Black, Heather. (2011). Connecting people
investments and business outcomes at Lowe’s, HR People + Strategy ,
34(2).
ConAgra Foods. (2010). Conagra foods recognized for its innovative
leadership development programs. Retrieved June 16, 2016, from http://
www.conagrafoods.com/investor-relations/news-ConAgra-Foods-
Recognized-for-Its-Innovative-Leadership-Development-Programs-
1486696
Coolen, Patrick & IJsselstein, Auke. (2015). A practitioner’s view on HR
Analytics. Retrieved July 18, 2016, from https://www.linkedin.com/pulse/
practitioners-view-hr-analytics-patrick-coolen
Crush, Peter. (2009, March). Engagement surveys: Gallup and best companies
face criticism. HR magazine.
Davenport, Thomas H., Harris, Jeane, & Shapiro, Jeremy. (2010). Competing on
talent analytics. Harvard Business Review.
 References 237

Drucker, P.F. (1954). The practice of management. New York: Harper & Row.
Fitz-enz., J. (1978). The measurement imperative. Personnel Journal, 54(4).
Friedman, T. (2014, February). How to get a job at Google. New York Times.
Gallup. (2012). State of the workforce report. Washington, DC: Gallup Inc.
Garvin, D.A., Wagonfel, A.B., & Kind, L. (2013). Google’s project oxygen: Do
managers matter (Case 313–110). Boston, MA: Harvard Business School.
Gates, S. (2002). Value at work: The risks and opportunities of human capital
measurement and reporting. New York: The Conference Board.
Gates, S. (2008). Strategic human capital measures: Orientation, accountability
and communication. New York: The Conference Board.
Gebicke, S., & Magid, S. (2012). Lessons from around the world: Benchmarking
performance in defense. MoG_Benchmarking, McKinsey & Company.
GoogleEDU (2012, July 5). School’s in session at Google. The Wall Street
Journal. Retrieved June 22, 2016, from http://www.wsj.com/articles/SB1
0001424052702303410404577466852658514144
Harrison, David A., Newman, Daniel A., & Roth, Philip L. (2006). How important
are job attitudes? Meta analytic comparisons of integrative behavioral
outcomes and time sequences. Academy of Management Journal, 49(2),
305–325
Hausknecht, J. (2013, December). They are watching you at work. The Atlantic.
Huselid, M.A. (1995). The impact of human resource management practices
turnover, productivity and corporate financial performance. Academy of
Management Journal, 38(3), 635–672.
Kirkpatrick, D.L. (1959a). Techniques for evaluating training programs. Journal
of American Society of Training Directors, 13(11), 3–9.
Kirkpatrick, D.L. (1959b). Techniques for evaluating training programs: Part
2—Learning. Journal of American Society of Training Directors, 13(12),
21–26.
Kirkpatrick, D.L. (1960a). Techniques for evaluating training programs: Part
3—Behavior. Journal of American Society of Training Directors, 14(1),
13–18.
Kirkpatrick, D.L. (1960b). Techniques for evaluating training programs:
Part 4—Results. Journal of American Society of Training Directors, 14(2),
28–32.
Kowske, B. (2012). Employee engagement: Market review, buyer’s guide,
provider profiles. Oakland, CA: Bersin by Deloitte.
Laney, D. (2001). 3D data management: Controlling data volume, velocity, and
variety. Stamford, CT: META Group Inc.
LaValle, S., Lesser, E., Shockley, R., Hopkins, M.S., & Kruschwitz, N. (2011).
Big data, analytics and path form insights to value. MIT Sloan Management
Review, 52(2).
Lavoie, A. (2014). Measure these 5 talent metrics for greater success in hiring and
managing employees. Entrepreneur. Retrieved July 18, 2016,from https://
www.entrepreneur.com/article/239117
238 Winning on HR Analytics

Lewis, M. (2003). Moneyball: The art of winning an unfair game. New York, NY:
W. W. Norton & Company.
Lombardi, M., & Laurano, M. (2012). Human capital management trends.
Boston, MA: Aberdeen Group.
Macy, John (2015). HR will never be strategic. Retrieved July 22, 2016
from https://www.linkedin.com/pulse/hr-never-strategic-john-macy?
trk=mp-reader-card
Mavrinac, S., & Siesfeld, T. (1998). Measures that matter: An exploratory
investigation of investor’s information needs and value priorities. London:
Ernst & Young Center for Business Innovation.
Michaels, E., Handfield-Jones, H., & Axelrod, B. (1997). War for talent. New
York, NY: McKinsey & Company.
Neilson, G., & Wulf, J. (2012, April). How many direct reports? Harvard
Business Review.
Organ, Dennis. (1988). Organization citizenship behavior, the good soldier
syndrome. Lexington, MA: Lexington Books
Palmer, S. (2010). Making the business case for learning and development:
5 steps for success. Chapel Hill, NC: UNC Kenan–Flagler Business
School.
Pease, G., Beresford, B., & Walker, L. (2014). Developing human capital: Using
analytics to plan and optimize your learning and development investments.
Hoboken, NJ: Wiley.
Penrose, E. T. (1959). The theory of the growth of the firm. New York: John
Wiley.
Phillips, J. J. (1996a). How much is the training worth? Training and Development,
50(4), 20–24.
PwC Saratoga. (2014). Trends in workforce analytics: Capturing the latest
benchmark results from US human capital effectiveness report. New York:
PwC. Retrieved June 16, 2016, from http://www.pwc.com/en_US/us/
hr-management/publications/assets/pwc-trends-in-workforce-analytics.
pdf
Rucci, Anthony J., Kirn, Steven P., & Quinn, Richard P. (1998). The
employee–customer–profit chain at Sears. Harvard Business Review
(January–February).
Schneider, C. (2006, February). The new human capital metrics. CFO.
Shobha, V. (2015, July 31). Sridhar Vembu, the code breaker. OPEN.
Sun Microsystems. (2014). Vestrics—Client success stories. Retrieved November
15, 2015, from http://www.vestrics.com/client-success-stories.html
Training Industry Inc. (2016). Size of the training Industry. Raleigh, NC: author.
Retrieved June 16, 2016, from https://www.trainingindustry.com/wiki/
entries/size-of-training-industry.aspx
Vickers, M. (2010). The crummy state of talent management metrics (and
what to do about it).i4cp. Retrieved July 23, 2016, from http://www.
 References 239

i4cp.com/trendwatchers/2010/06/02/the-crummy-state-of-talent-
management-metrics-and-what-to-do-about-it
Waterman, R. H., Peters, T. J., & Phillips, J. R. (1980). Structure is not organization.
Business Horizons, 23(3), 14–26.
 Index

Aberdeen Group 17 classroom based development 85

activity-based compensation 45 clustering 122
analytics companies 79 cognitive analytics 8
analytics in workforce or people communication skills 130, 160
management 5 company executives and HR
artificial neural networks 122 function 4
attrition analysis 121 compensation 133
automobile company 27 annual merit increase process 140
business levers of 134
badges 29
CTC statement, valuing benefits
Barney, J. 2
using 147
Bayesian Theorem 89
high performance organization
Beane, Billy 5
144
benefits
organization structure and cost of
business levers of 134
management 135
portfolio management of 148
traditional measures of 139
tailoring variable pay 151
competencies 124
big data 10
competency baselines 125, 128
Boudreau 11
leadership development 130
business acumen 154
talent acquisition 131
business outcomes 16
competency baselines 125, 128
Cascio 11 competency index 129
CedarStone 17 competing on talent analytics 5
centralized structure 157 competitive advantage integration 12
Centre of excellence (CoE) 157 ConAgra 29, 97
china wall for HR 11 conference board survey 15, 23
classification methods 122 context based search 79
 Index 241

controlled group 89 Fortune 5

Coolen, Patrick 155 Foundations of Leadership (FoL)
corporate world 5 Program 97
cost of hiring 77 functional structure 158
cost of management 135 fundamental analytic error 167
cost of quality 129 fundamental attribution error 167
CTC statement, valuing benefits fun events 41
using 147
culture 159 Gates, Stephen 25
Google 4, 5, 8, 45, 97, 101
data analysis 155, 159 grade point scores 45
data interpretation skills 155 Graicunas, V.A. 49
data presentation skills 155 Green, David 165
dealer satisfaction score (DSS) 28
Harvard Business Review 5
decentralized structure 157
Hausknecht, John 6
deep and wide approach 31
high performance culture 43
descriptive statistical techniques
high performance work systems 15
121
High Productivity 25
Drucker, Peter 1
hiring specification validity 73
employee-customer-profit chain horizontal alignment 3
99 HR analytics team leader 156
employee engagement HR domain skills 155
attrition analysis, predictive Human Capital Lab 97
modelling for 121 human capital management
business levers of 98, 102 (HCM) 16
employee retention 106 Human Dynamics Laboratory 29
engagement curve 112 human factor 10
lead indicators 107 human resources accounting and
and predicting 111 utility analysis 5
employee referrals 112 human resources (HR) 6
LTM 104 alignment 3
measuring attrition 102 building the cube 37
modeling for surveys 108 business outcomes
year to date (YTD) 105 impact on 16
employee growth 57 linkage to 16
employee surveys 60 measure linkage of 24
employee welfare costs 133 business results linkages,
Employers of Choice 33 measuring 18
Ernst & Young Center for Business competitive advantage integration
Innovation 15 12
deep and wide approach 31
financial skills 155 evolution 3
Forbes 5 going from opinion to insight 40
formal methods 85 human capital alone 2
242 Winning on HR Analytics

impact, industry examples 29 non-controlled group 89

research evidence 23
stradegy validation framework ordinary least square (OLS)
43 regression 122
sudden interest in 7 organization structure 47
tryst with competitive business levers 47
advantage 1 compensation 135
Human Resources Management headcount in offices 59
(HRM) organization demographics and
data 159 succession planning 60
executing transformation 162 organization shaping and
people analytics 164 employee growth 56
skills 154 organization shaping through
staff 155 pyramid ratios 52
strategy 158 softer aspects, measuring 60
structure 157 traditional measures 48
support 158 Palmer 86
systems 158 Penrose, E.T. 2
Huselid 15 People Analytics 6, 164
IJsselstein, Auke 155 people capability maturity model 124
informal methods 85 Personnel Journal 4
Infosys 134 Philips, Jack 86
process control 81
Journal of ASTD 86 Project Oxygen 5

Key Performance Indicators quality of hire 75

(KPIs) 27
Kirkpatrick, Donald 86 regression analysis 121
research methodology skills 155
Laney 10 resource-based view (RBV) theory 2
last 12 months (LTM) 104 retention
Lavoie 94 attrition analysis, predictive
leadership 158 modelling for 121
lead indicators 107 fine-tune action items for 113
and predicting 111 predictability, building up for 116
employee referrals 112 return on investment (RoI) 27, 86, 91
logistic regression 122
sabermetrics 5
map linking 94 Silicon Valley 100
Marriott Vacation Club 23 skills, HRM 154
metrics value chain 8 Sociometry 29
microscope 10 software developer engagement 26
Microsoft 4, 98 Software Engineering Institute 124
Moneyball 5 statistical modeling technique 101
Moodrings 29 statistical skills 154
 Index 243

storification 161 methods 85

stradegy validation framework 43 return on investment 86, 91
Sun Microsystems 97 strategic alignment, right metrics
survival analysis 122 and measures for 93
Sysco Corp 45, 151 talent engagement 98
Talent Management Technology
talent acquisition (TA) 64 Adoption 18
business levers of 64 Tata Motors 133, 134
competencies 131 TCS 134
cost of hiring 77 Tech index 129
cycle time, opportunity cost of 72 telescope 10
effectiveness measures 68 The Atlantic 6
effort 67 trailing 12 months (TTM) 104
emerging measures 71
hiring specification validity 73 US Army 4
measuring and improving process
capability 81 vertical alignment 3
predictctable joining and Vickers 94
performance 78 visualization 161
quality of hire 75
Wall Street Journal 5
speed 66
workforce management 6
talent development 84
workforce science 6, 7
ABC Company 87
Base level and optimization Xerox Services HR analytics 29
level 87
four levels of the model 86 year to date (YTD) 105
 About the Authors

Ramesh Soundararajan is an HR professional with 25 years of

experience as a practitioner and consultant. An electrical engineer
from National Institute of Technology (NIT), Kozhikode, Ramesh
completed his master’s in personnel management and industrial
relations from Xavier School of Management (XLRI), Jamshedpur.
Crompton Greaves, TVS Whirlpool, Infosys Technologies
(more than a decade), and Sasken Communications are the promi-
nent companies he has worked with. Presently he is the founding
partner of Culstran LLP, a firm focused on consulting corporates
in the areas of culture, strategy, and analytics. He has worked as a
location head of HR, head of a CoE, as well as the head of the func-
tion. He has consulted with clients in India as well as in the USA.
In all his roles, he has pioneered an analytical approach to
reviewing information, integrating insights from across differ-
ent functions to help the function put its best foot forward. The
approach encompasses all HR domains such as performance man-
agement, learning and development, and talent acquisition and
retention. His blog on analytics, “HR3by2”, is widely referred to.
He is working with large corporations from developing analytics
competency in HR to designing interactive dashboards. He also
works with start-ups in the HR analytics space. He is a trained
assessor using the PCMM and CII HR models.

Dr Kuldeep Singh is Director, HR, at Capgemini, Bangalore. He

has 20+ years of experience in HR having worked in manufacturing
and IT sector. He has held various positions such as corporate head
(HR), strategic HR advisor, global head (OD), etc., for organizations
with headcount size from 4,500 to 67,000+. His last position was
corporate head (HR) and strategic HR advisor for UST Global Inc.
Prior to UST Global, he has worked at Infosys as senior manager and
head Performance Management System (PMS), spearheading PMS
and Organization Development (OD) interventions globally.
 About the Authors 245

Before his corporate career, he worked as an associate professor

(HR) at IIM Indore. He has specialized in HR/people analytics and
conducted open workshops on HR analytics with IIMs at metro
locations and in-house workshops for business partner HR teams
of a tier-I Indian technology services company. His publications on
HR analytics on social media, and in HR magazines and leading
business newspapers have been globally acclaimed. He is fortified
with HR and business analytics certifications from Wharton and
UC Berkeley and a PhD in HR from XLRI.