Unit 5

HR Analytics
• Human Resource analytics (HR Analytics) is defined as
the area in the field of analytics that deals with people
analysis and applying analytical process to the human
capital within the organization to improve employee
performance and improving employee retention.
• HR analytics doesn’t collect data about how your
employees are performing at work, instead, its sole aim
is to provide better insight into each of the human
resource processes, gathering related data and then
using this data to make informed decisions on how to
improve these processes.
The 4 types of HR analytics
1. Descriptive:
2. PTO(paid time off days)for an year
a. Employee Turnover
3. Diagnostic:
a. Employee Absenteeism
b. Employee Engagement
4. Predictive:
a. Recruitment
b. Retention
5. Prescriptive:
a. Staffing
b. Attrition
Employee attrition vs turnover vs
churn
• Employee attrition :An employee’s departure is considered attrition if
it meets the following criteria:
a. The departure is voluntary.
b. The company is not rehiring or re-filling the position.
• Employee turnover: It is the percentage of employees that:
a. leave your company after a certain period of time AND
b. That you intend to refill the position
• Employee turnover can be Voluntary & Involuntary
• Employee churn: refers to the total number of attrition and turnover
combined.
Top 5 HR Analytics
Top 5 types of HR Analytics Every Human Resource Manager Should
Know
• It goes without saying, that employees are an asset and vital to the
success of any organization.
1. Employee churn
2. Capability
3. Organizational culture
4. Capacity
5. Leadership
Employee churn in python
#Import modules
import pandas # for dataframes
import matplotlib.pyplot as plt # for plotting graphs
import seaborn as sns # for plotting graphs
• % matplotlib inline
• Loading dataset:
• Data=pandas.read_csv(‘HR_comma_sep.csv’)
• Data.head()
You can check attributes names and datatypes using info().
• Data.info()
• This dataset has 14,999 samples, and 10 attributes(6 integer, 2 float, and 2
objects).
• No variable column has null/missing values.
• You can describe 10 attributes in detail as:
• satisfaction_level: It is employee satisfaction point, which ranges from 0-1.
• last_evaluation: It is evaluated performance by the employer, which also ranges
from 0-1.
• number_projects: How many numbers of projects assigned to an employee?
• Average_monthly_hours: How many average numbers of hours worked by an
employee in a month?
• Time_spent_company: time_spent_company means employee experience. The
number of years spent by an employee in the company.
Work_accident: Whether an employee has had a work accident or not.
Promotion_last_5years: Whether an employee has had a promotion in
the last 5 years or not.
Departments: Employee’s working department/division.
Salary: Salary level of the employee such as low, medium and high.
• Left: Whether the employee has left the company or not.
Let’s Jump into Data Insights
• In the given dataset, you have two types of employee one who stayed
and another who left the company. So, you can divide data into two
groups and compare their characteristics.
• Here, you can find the average of both the groups using groupby() and
mean() function.
• Left = data.groupby(‘left’)
• Left.mean()
• Data.describe()
Data Visualization
Employees Left:
Let’s check how many employees were left?
Here, you can plot a bar graph using Matplotlib.
Left_count=data.groupby(‘left’).count()
plt.bar(left_count.index.values, left_count[‘satisfaction_level’])
plt.xlabel(‘Employees Left Company’)
plt.ylabel(‘Number of Employees’)
• plt.show()
Data.left.value_counts()
0 11428
1 3571
• Name: left, dtype: int64
• Similarly, you can also plot a bar graph to count the number of
employees deployed on How many projects?
Building a Prediction Model(Pre-Processing Data):
In order to encode this data, you could map each value to a number.
E.g. Salary column’s value can be represented as low:0, medium:1, and
high:2.
• This process is known as label encoding.
• Import LabelEncoder
• from sklearn import preprocessing
• #creating labelEncoder
• le = preprocessing.LabelEncoder()
• # Converting string labels into numbers.
• Data[‘salary’]=le.fit_transform(data[‘salary’])
data[‘Departments ‘]=le.fit_transform(data[‘Departments ‘])
#Spliting data
X=data[[‘satisfaction_level’, ‘last_evaluation’, ‘number_project’,
‘average_montly_hours’, ‘time_spend_company’, ‘Work_accident’,
‘promotion_last_5years’, ‘Departments ‘, ‘salary’]]
• y=data[‘left’]
• #Import train_test_split function
• from sklearn.model_selection import train_test_split
• # Split dataset into training set and test set
• X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3,
random_state = 100)
# Building Model using Gradient Boosting Classifier model
from sklearn.ensemble import GradientBoostingClassifier
#Create Gradient Boosting Classifier
gb = GradientBoostingClassifier()
#Train the model using the training sets
gb.fit(X_train, y_train)
#Predict the response for test dataset
• y_pred = gb.predict(X_test)
#Evaluating Model Performance
#Import scikit-learn metrics module for accuracy calculation
from sklearn import metrics
# Model Accuracy, how often is the classifier correct?
Print(“Accuracy:”,metrics.accuracy_score(y_test, y_pred))
# Model Precision
print(“Precision:”,metrics.precision_score(y_test, y_pred))
# Model Recall
• print(“Recall:”,metrics.recall_score(y_test, y_pred))
Employee Attrition in python
Import pandas as pd
import numpy as np
• import matplotlib.pyplot as plt
• we have 1470 rows and 35 columns in our dataset.
• Attrition_dataset = pd.read_csv(r”C:\Datasets\
employee_attrition_dataset.csv”)
• print(“Dataset rows and columns:”, attrition_dataset.shape)
• attrition_dataset.head()
Exploratory Data Analysis
• To identify missing values
• Attrition_dataset.isna().sum()
• Pie chart:
• Attrition_dataset.Attrition.value_counts().plot(kind=‘pie’,
autopct = ‘%1.0f%%’, figsize=(8, 6))
Next, let’s see how the employee attrition ratio varies with the marital
status of an employee.
Attrition_dataset.groupby([‘MaritalStatus’,‘Attrition’]).size().unstack().pl
ot(kind=‘bar’, stacked=True, figsize=(8, 6))
• The output below shows that the attrition rate is the highest among
employees with single marital statuses.
The following script shows the employee attrition rates among different
age groups.
• Attrition_dataset.groupby([‘Age’, Attrition’]).size().unstack().plot
(kind=‘bar’, stacked=True, figsize=(12, 8))

The output shows that, in

our dataset, employee
attrition rates are higher
among employees aged less
than 35. The attrition rates
are zero among the
employees aged 59 and 60.
Data Preprocessing
Feature_set = attrition_dataset.drop([‘Attrition’], axis=1)
• labels = attrition_dataset.filter([‘Attrition’], axis=1)
• Feature_set.dtypes

Machine learning algorithms

work with numbers, but our
feature set contains some non-
numeric columns, as you can
see from the output of the
following script.
We need to convert the non-numeric columns in our dataset to
numeric columns. Let’s do that now.
The following script separates categorical features from numeric
features in our dataset.
Cat_col_names = [‘BusinessTravel’, ‘Department’, ‘EducationField’,
‘Gender’, ‘JobRole’, ‘MaritalStatus’, ‘Over18’, ‘OverTime’ ]
num_cols = feature_set.drop(cat_col_names, axis=1)
• cat_columns = feature_set.filter(cat_col_names, axis = 1)
• You can use the one-hot encoding approach to convert categorical
features to numeric features. The following script uses the Pandas
get_dummies() method to convert categorical features in our dataset
to one-hot encoded numeric features. The output shows the total
number of one-hot encoded columns.
• Cat_columns_one_hot = pd.get_dummies(cat_columns,
drop_first=True)
• cat_columns_one_hot.shape
• Output: (1470, 21)
Finally, you can concatenate the default numeric features with the one-
hot encoded numeric features to form the final feature set.
X = pd.concat([num_cols,cat_columns_one_hot], axis=1)
• X.shape
• Output:
• (1470, 47)
Similarly, we can convert the simple “Yes” and “No” values from our labels set to binary 1 and
0 values, respectively, using the following script:
y = labels[‘Attrition’].map({‘Yes’: 1, ‘No’: 0})
y.head()
Output:
0 1
1 0
2 1
3 0
4 0
• Name: Attrition, dtype: int64
Model Training and Predictions
From sklearn.model_selection import train_test_split
• X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.20,
random_state = 42)
• From sklearn.ensemble import RandomForestClassifier
• rf_clf = RandomForestClassifier(n_estimators = 40, random_state =
42)
• rf_clf.fit(X_train, y_train)
• pred = rf_clf.predict(X_test)
From sklearn.metrics import classification_report, accuracy_score
print(classification_report(y_test,pred ))
• print(accuracy_score(y_test, pred ))
• The output shows that the model achieves an accuracy of 87.07% on
the test set.
• As a last step, we want to see what the most important features an
organizations must consider to avoid employee attrition. To do so, you
can use the feature_importances_ attribute from the trained random
forest classifier.
• Important_features = pd.Series(rf_clf.feature_importances_,
index=X.columns)
• important_features.nlargest(10).plot(kind=‘barh’)
The above output
shows that monthly
income and overtime
are the most
important reasons
behind employee
attrition.
• Capability Analytics:
• Undoubtedly, the success of any business to an extent depends on the
level of expertise of the employees and their skills.
• Organizational Culture Analytics:
• Organizational culture analytics is a process of assessing and
understanding better the culture at your workplace.
• Capacity Analytics:
• The aim of capacity analytics is to establish how operationally
efficient is your workforce.
• Leadership analytics:
• Leadership analytics analyzes and unpacks various aspects of
leadership performance at a workplace to uncover the good, bad and
the ugly!
• Data can be collected through qualitative research and quantitative
research by using a mix of both methods like surveys, polls, focus
groups or ethnographic research.
HR Metrics Dashboard
The HR Metrics dashboard is an important part of Human Resource
planning and strategy. Here are the top 3 functions of an HR dashboard:
1. To monitor human capital
2. Help HR perform better
3. Tackle problem areas
Predictive HR Analytics Trends –
2020
• Here are some HR Analytics trends that will be big in 2020:
• 1. AI and automation in hiring: With the advent of Artificial Intelligence (AI) and
automation in HR, recruitment will go a sea of change.
2. Virtual onboarding and training: Induction programs and training will become
more virtual and interactive. This will be primarily due to massive strides in AI and
Natural Language Processing (NLP) and Machine Learning (ML).
3. Performance management: Performance reviews and management will get
more detailed and personal due to detailed analytics that HR Analytics tools
provide.
4. Predictive reports: Rich workforce analytics will help in identifying attrition risks
and preemptively taking measures to arrest that. Not only that, it will help in
identifying the strengths and weaknesses of employees helping in designing better
and efficient teams.
• Benefits of HR Analytics:
• 1. Improve your hiring process
2. Reduce attrition
3. Improve employee experience
4. Make workforce productive
5. Improve your talent processes
6. Gain employee trust
Payroll analytics
• Several large businesses use payroll analytics to improve profitability.
• Attendance: Measuring absenteeism is one of the most common
payroll analyses performed. An in-depth analysis might provide
significant information.
• Budget: Comparing your actual labour costs to your budgeted labour
costs can reveal problems in several critical areas.
• Turnover: A high turnover rate often signals low employee morale.
Are your pay rates competitive with those of similar companies?
• Alignment: Alignment is the process of matching payroll costs to
company goals.
The strategic HR insights that can be extracted using payroll analytics
include:
Minimize/Eliminate Errors:
• For most firms, payroll is the single largest expense head.
• For some large MNCs, it can account for 50% or more of the
company’s overheads.
• In such a scenario, even small errors can snowball into major
problems.
• Identifying the cause and frequency of errors can help organizations
devise remedial action and save operational costs.
Informed Business Decisions:
• The analysis of payroll performance helps generate accurate annual
forecasts.
• Accurate forecasts enable efficient budget and cash flow management,
which is particularly important when the business is going through a
period of change or growth.
• Planning for Future Growth:
• The correct analysis of payroll data can also support business leaders
in formulating strategies for future growth.
• Payroll analytics could also arm you with data to decide on business
expansion in a particular location over another based on the cost of
compensation, training cost, tax liabilities, etc.
Superior Hiring Decisions:
• An assessment of the challenges of deploying a larger workforce to
power growth should accompany any planning for future growth.
• When the business is spread across geographies, accurately
evaluating the costs and risks of employing staff in multiple locations
is vital for the planning and hiring process.
• Improved Employee Retention:
• Payroll analytics can reveal the correlation between monetary
compensation and performance at work.
• Uncovering such insights from HR payroll helps draft employment
contracts that help attract top talent and retain them.
Marketing analytics
• By applying technology and analytical processes to marketing-related data,
businesses can understand what drives consumer actions, refine their
marketing campaigns and optimize their return on investment.
• What can you do with marketing analytics?
• 1. How are our marketing activities performing today? How about in the
long run? What can we do to improve them?
• 2. How do our marketing activities compare with our competitors? Where
are they spending their marketing dollars? Are they using channels that we
aren’t using?
3. What should we do next? Are our marketing resources properly allocated?
Are we devoting time and money to the right channels? How should we
prioritize our investments over a certain time period?
 How marketing analytics works
• 1. Identify what you want to measure
• 2. Use a balanced assortment of analytic techniques and tools
• a. Analyze the present
• b. Predict or influence the future
• 3. Assess your analytic capabilities, and fill in the gaps
• 4. Act on what you learn
Some popular analytics models and methods include:
Media Mix Models (MMM): Attribution models that look at aggregate
data over a long period of time.
Multi-Touch Attribution (MTA): Attribution models that provide person-
level data from across the buyer’s journey.
• Unified Marketing Measurement (UMM): A form of measurement
that integrates various attribution models including MMM and MTA
into comprehensive engagement metrics.
How Organizations Use Marketing
Analytics
• Using this data, your team can gain insights into the following:
• Product Intelligence: Product intelligence involves taking a deep dive into
the brand’s products as well as analysing how those products stack up
within the market.
• Customer Trends and Preferences: Which products are they buying and
which have they researched in the past? Which ads are leading to
conversions and which are ignored?
• Product Development Trends: Analytics can also offer insight into the types
of product features consumers want.
• Customer Support: Analytics also helps uncover areas of the buyer’s
journey that could be simplified or improved.
• Messaging and Media: In addition to traditional marketing channels such as
print, television and broadcast, marketers must also know which digital
channels and social media networks consumers prefer. Analytics answers
these key questions:
• What media should you be buying?
• Which are driving the most sales?
• What message is resonating with your audience?
• Competition: How do your marketing efforts compare with the competition?
How can you close that gap if there is one? Are there opportunities your
competitors are capitalizing on that you may have missed?
• Predict Future Results: If you have a thorough understanding of why a
campaign worked, you’ll be able to apply that knowledge to future
campaigns for increased ROI.
Challenges
• Data Quantity: Big data
• Data Quality
• Lack of Data Scientists
• Selecting Attribution Models: Determining the model that provides
the right insights can be tricky.
• Correlating Data: In this same vein, because marketers are collecting
data from so many different sources, they must find a way to
normalize it to make it comparable.
BIG DATA AND MARKETING
• Big data has created big opportunities for marketing departments all over the
globe.
• The easy access to vast amount of information based on customer interactions
has enabled marketers to go from knowing the customer as a demographic to
understanding him or her as an individual.
• Bessen (1993) argues that successful marketing always has relied on a two-
way information flow between the marketer and the customer.
• The marketers’ challenge has always been to collect detailed demographic and
life-style information on consumers that can be used as a basis for effective
and efficient activities.
• Consequently, one of the greatest challenges for marketing departments is to
create an information ecosystem by combining patterns of data from different
sources.
• Online giants such as Google and Facebook have made harnessing the data on
the web into extremely lucrative businesses.
• Based on big data, Google and Facebook can now offer their clients tailored
campaigns based on for example online searches, posts and messages.
• Marketers involved with Internet marketing know the potential reward of cost-
efficient tools such as social media feeds, user videos and image posts etc.
• All the online tools have converged into massive dashboards that give
marketers real-time and holistic views of the consumers and their ongoing
activities (Smith, 2013).
• Marketers have been armed with advertising optimization capabilities overall,
based on the growing amount of big data available to them.
• THE CHALLENGES OF BIG DATA FOR MARKETERS
• 1.Prajicek (2013) criticizes many marketing departments for committing that
the data “more is better”-fallacy.
• 2. Another common mistake marketers commit is that they spend too much
time “in the cloud” looking at macro trends.
• 3. Great data and great insights are often wasted when big data is not being
leveraged by little data and interpreted based on the integrations and
relationships it’s based on (Ross, 2013; Fox & Do, 2013).
• 4. Another challenge of big data stems from the established clash between
marketing and sales. Big data can erode long-term marketing commitments in
order to promote short-term sales.
• 5. One of the main criticisms against the use of big data in marketing is the
implication on individual privacy. How much information on about consumers
can big corporation store before it becomes an intrusion on private life?
7 Ways Brands Can Use
Personalization To Create Unique
Online Shopping Experiences
• Personalized online shopping experiences refer to ways customers interact with
your brand that’s unique to their profile or preferences.
• Here’s how a personalization strategy can benefit your brand:
• improving customer experience
• Create consistency across channels
1. Inform future business decisions
2. Reduce support debt: Integrated apps can make returns and exchanges
painless for online shoppers and lessen the burden on your support team.
3. Increase customer loyalty
4. Optimize conversion rate: converting more browsers to buyers.
5. But beware: Consumers are over three times as likely to avoid brands that
over-personalize versus those that don’t personalize at all.
7 ways to delight customers with
personalized online experiences
• Here are seven ways they’re creating unique shopping experiences for
their customers:
• Build a virtual fitting room
• Offer personalized product recommendations
• Create customized experiences and products
• Upgrade to smart, relevant reviews
• Offer tailored product education
• Mirror the in-store experience online
1. Close the personalization loop with returns.
Role Of Big Data Analytics In Digital
Marketing Strategy
Big data is crucial in digital marketing because it provides companies
with deep insights about consumer behavior.
• Google is an excellent example of big data analytics in action. Google
leverages big data to deduce what consumers want based on a variety
of characteristics, such as search history, geography and trending
topics.
• Big data mining has resulted in Google’s secret source of proactive or
predictive marketing: determining what consumers desire and how to
incorporate that knowledge into the company’s ad and product
experiences.
But your company doesn’t have to be a tech giant to use big data
analytics successfully. Here are four key ways companies of all sizes can
benefit from big data:
• 1.Receiving Data Analysis In Real Time
• 2. Enabling data Targeted Advertising
• 3. Analyzing Customer Insights
• 4. Creating Relevant Content
• 5. Protecting Customer Privacy
What is Data Monetization?
• Data monetization is using your data to add to or increase your
revenue stream. Data monetization can be achieved through both
internal and external methods:
• Internal data monetization is the method of using data and analytics
to make informed business decisions that turn into measurable
improvements to the way a company does business.
• Examples: upselling or cross selling opportunities with targeted
marketing, supply chain planning to improve inventory and
transportation management, or hiring and retaining staff based on
data and analytics.
How to Identify Opportunities for Internal Data Monetization
• To take advantage of data monetization internally, start by reviewing
your data strategy—because anytime you take a holistic look at the role
data plays in your company, you will uncover areas for improvement and
optimization.
• Next, look at the capabilities of your data stack—are you taking
advantage of everything your technology has to offer, or do you need to
consider new technology and infrastructure to get more from your data?
• Make sure you are considering your whole business, from tools and
resources to people and processes.
• External data monetization is the method of creating a product or
service using your internal data assets and selling them to a third
party.
• Examples: creating benchmark reports; selling lists of contact
information, or verified address and property data; or providing
access to academic research or survey results behind a paywall.
What are the Different Ways You Can Monetize Data Externally?
• There are two ways to monetize your data externally—either by
1. offering data as service or product - Data as a Service provides data
either as a one-time product or as a subscription service to a data
set that is constantly updated.
2. by offering insights as a service or product - Insights as a Service
provides analytics and expertise attached to your data that you can
either sell as a one-time product or as a subscription-based service.
Machine-to-Machine Analytics
• Machine-to-Machine (M2M) means that no human intervention is required
whilst devices are communicating with each other.
• It refers to technologies that allow both wireless and wired systems to
communicate with other like devices.
• M2M is considered an integral part of Internet of Things (IoT) and brings
several benefits to the industry and business.
• M2M, is a broad term that refers to any technology that allows networked
devices to exchange information and perform actions without the need for
human intervention.
• Artificial intelligence (AI) and machine learning (ML) facilitate system
communication and allow them to make autonomous decisions.
• M2M communication is commonly used to refer to cellular communication
for embedded devices. ATMs, for example, may be authorized to dispense
a specific amount.
• primary goal of machine-to-machine technology is to collect sensor data
and send it to a network.
• Unlike SCADA or other remote monitoring tools, M2M systems frequently
use public networks and access methods, such as cellular or Ethernet, to
reduce costs.
• Sensors and RFID, a Wi-Fi or cellular communications link, and autonomic
computing software programmed to assist a network device in interpreting
data and making decisions are the main components of an M2M system.
• These M2M applications translate data into preprogrammed, automated
actions.
Types of Machine to Machine
Communication
• M2M communication can be classified into two types:
• 1. Wired communication:
• The data transfer between devices in wired M2M communication
takes place over a wired transfer medium.
• Fiber optic cables, EtherCAT, or even coaxial cables can be used.
• Communication between devices in the same LAN network connected
only by wires is also included in wired M2M communication.
• 2. Wireless M2M Communication:
• Wireless M2M communication occurs when machines use wireless
communication methodologies to communicate between devices.
• The Internet of Things, or IoT, refers to M2M communication that
primarily uses a wireless network for communication.
• Wireless communication methodologies used range from radio waves to
the most recent 5G technology.
• RFID, or Radio Frequency Identification, is an ancient technology that is
now widely used in the logistics industry to track a package from
beginning to end. It is also used in blockchains for provenance technology.
• Near-Field Communication (NFC) is similar to RFID, but it can only be used
for short-range data transfer. It’s common in access control and payment
systems.
• WiFi, or wireless fidelity, is a popular method of wireless internet
access in homes and offices.
• WiFi 6e, the most recent WiFi technology approved by the FCC, has
the potential to significantly improve IIoT communication.
• Cellular networks can be used for wireless M2M communication as
well.
• The First wireless credit card readers communicated using GSM or
other 2G technologies.
• Over a 5G network, thousands of devices would be able to ‘talk to
each other’ very quickly and with very little lag.
• The implementation of 5G in the industrial setting is said to be the
beginning of a new industrial revolution.
Four Categories of Analytics that
Draw Insights from Product Data
• There are four categories of analytics that can be used to draw insight
out of PLM(product life cycle) data.
• First, descriptive analytics are the easiest to execute and involve
using data to explore trends and performance after they happen.
Most organizations already engage in forms of descriptive analytics
with their PLM product data today.
• Diving deeper, the next level of analysis involves diagnostic analytics,
which explain the reasons behind historical trends. This information
can provide insights about which variables are most responsible for
an outcome.
• The third category is predictive analytics, which uses advanced
machine learning algorithms to peer into the future and understand
the most likely outcomes before they happen. For example, imagine
an algorithm that could predict that a product in early stages of
development will likely fall short of market expectations. Knowing this
in advance through predictive analytics would allow resources to be
diverted to R&D products that are more likely to be successful.
• Finally, the fourth and most advanced category is prescriptive
analytics, which makes evidence-based recommendations to help
optimize future outcomes. Imagine asking your data to tell you how
to decrease product cycle times, or which suppliers to use to meet
demand needs on time.
• The analytics offered by true PLI (product lifecycle intelligence)can
take PLM (product lifecycle management) data and use it to create
models that can guide decisions to maximize value. As an organization
advances from descriptive to prescriptive, the potential for results-
driven insights, particularly before critical investment decisions are
made, increases sharply. PLI helps organizations predict the impact of
product development decisions on key business performance metrics,
like demand, cycle time, cost, quality, regulatory compliance,
manufacturability and supply chain efficiency.