Because learning changes

everything. ®

Chapter 01:
Introduction to
Marketing
Analytics

Part 01: Overview of

Marketing Analytics and
Data Management

Copyright © 2025 by McGraw Hill Education (India) Private Limited.

 Introduction to Marketing Analytics

How does Expedia, Orbitz, or How does Stitch Fix achieve the
Hotels.com determine the price highest-ever rate of purchased
to quote when you are shopping items per “Fix” for its female
for a hotel room? customers?

How does Chat GPT write an

How does Spotify know what
essay about marketing analytics
songs to suggest for you?
in a few seconds?

2
 Marketing Analytics Defined

Marketing
analytics uses
data, statistics,
mathematics, and
technology to
solve marketing
business
problems.
Modeling and
software drive
marketing
decisions.
The fastest growing
field of analytics
applications.
Increasingly
applied, and the
impact and benefits Access the text alternative for this image.

are evident. Source: Google Trends. 3

Analytics Levels and Their Impact on Competitive Advantage

As more advanced techniques emerge, higher data management and

analysis maturity are required for a competitive advantage.

Access the text alternative for this image.

Source: Adapted from SAS. 4

 Analytics Levels

Descriptive analytics are used to explain or quantify the past.

• Data queries, visual reports, descriptive statistics.
Predictive analytics use models of past data to explain the
future.
• For example, historic sales can predict future sales.
Prescriptive analytics identifies the optimal course of action
or decision.
• UPS route optimization, Amazon’s price optimization.
Artificial Intelligence (AI) and cognitive analytics are
designed to mimic human-like intelligence for certain tasks, like
discovering patterns.
• This type of analytics uses machine learning to understand
new data.
• Hitachi uses AI to discover patterns typically undetected by
humans.
5
 Defining the Right Business Problems

Understanding requires deep

knowledge of the customer’s
path. How do you arrive at the right
• How they search. business problem?
• Where they purchase. • Understand the intent
behind the question.
• Their satisfaction.
• Include stakeholder input
Problem identification
through discovery
uncovers strategic business
methods.
opportunities.
• Discovery begins with:
• To improve market share.
what, who, where, when,
• To establish a better why, and how.
customer relationship.
• To position the company to
take advantage of
innovation. 6
 The SMART Principles

The SMART principles can be a goal-setting technique.

Equally as important as following the SMART principles is examining:

• The potential success of the analytics project.
• Whether it makes a valuable impact.
Access the text alternative for this image.

7
 Data Sources

Data consists of both primary Sources of secondary data:

and secondary data. • Public datasets.
• Primary data is collected • Online sites.
for a specific purpose.
• Mobile data.
• Secondary data relies on
existing data collected for • Channel partners.
another purpose. • Commercial brokers.
• Corporate information.
• Government sources.

8
 Types of Data

Structured data is made up Unstructured data includes

of records organized in rows text, images, videos, and
and columns. sensor data.
• Can be stored in a • No defined structure.
database or in a • Content does not fit into a
spreadsheet formula. table format.
• Includes numbers, dates, • Requires advance analytics
and text strings. to prepare and analyze.
• Easy to access and • Technology has advanced
analyze. to support manipulation
and exploration of this
data.

9
 Data Measurement

Numerical data can be discrete Categorical data may be:

(integer) or continuous. • Binary – can have two values.
• Discrete data – whole • Nominal – has no meaningful order.
numbers.
• Ordinal – has meaningful values.
• Continuous data may include
decimals.

10
 Metric Measurement Scales

Metric scales can be measured as intervals or ratios.

• Both scales possess meaningful, constant units of measure.
• The distance between each point of the scale are equal.
However, there is a difference between these scales.
• Interval variables do not include an absolute zero.
• Ratio scales have an absolute zero point and can be discussed in
terms of multiples.

11
 Predictors versus Target Variable

Variables are Does weather impact ice cream sales?

characteristics or
• Weather is the independent variable.
features that pertain
to a person, place, • It influences or drives the dependent, target,
or object. or outcome variable which is ice cream sales.

12
 Modeling Types:
Supervised versus Unsupervised Learning

Supervised learning Unsupervised learning has

suggests the target variable is no previously defined target
known. variable.
• A training dataset helps • The goal is to model data
“learn” the relationship. to discover and confirm
• A validation dataset patterns.
assesses the algorithm’s This technique may include:
accuracy. • Association analysis
• A testing dataset offering product
evaluates the final selected suggestions based on past
algorithm. purchases.
• The algorithm is applied to • Cluster analysis which
new, unlabeled data. groups customers based on
If the target variable is key variables.
continuous, results are a
prediction.
13
Exhibit 1-9: Supervised Learning Steps

Access the text alternative for this image.

14
 The 7-Step Marketing Analytics Process

• The 7-step marketing analytics process is iterative and continuously

evolves to develop and manage improvements in the modeling cycle.
• Each step plays an important role in achieving a successful outcome.
Access the text alternative for this image.

15
 Step 1: Business Problem Understanding

Most marketing analytics models are Exactly what are

developed when a business identifies a you trying to
problem. understand and
• Develop a model using analytics to solve?
better understand the problem and How will the
design a solution. stakeholder(s)
• A key element is to question whether use the results?
the problem is the correct problem. Who will be
affected by the
results?
Is this a single,
short-term
problem or an
ongoing
situation?

16
 Step 2: Data Understanding and Collection

The analyst must identify where data is stored, its format, and how it can
be combined to understand the question.
• Examine databases, interview stakeholders, and observe processes
to confirm the identified problem is the actual problem.
Once the problem is understood, the analyst samples data from
databases for records to analyze.
• For example, examining past purchases and customer returns.
Marketing analysts must have a good understanding of the types and
sources of data.
• The data’s origin may directly affect the decision.

17
 Step 3: Data Preparation and Feature Selection

Data in different formats is combined in this step.

• Identify the unit of analysis; the target and the predictor variables.
• Examine target and predictor data columns both visually and
statistically.
• Clean the data – deal with missing values, data errors, and outliers.
• Merge data from different sources so data is measured consistently
and then used to develop the models.

Other features are further refined in this step.

• Adjusting date formats.
• Increase accuracy by including predictors with a strong target variable
relationship – some may be eliminated or transformed.
• Understanding the meaning of each variable and its unit of analysis is
essential in this step.

18
 Step 4: Modeling Development

In this step, the analyst selects the method to use.

• Choice depends on the target variable and problem.
• Options – classification, prediction, clustering, or association.
If the problem is unsupervised, the analyst partitions data into datasets.
• Training, validation, and testing.
The analyst should decide on appropriate modeling techniques.

Different models should be tried to find the one providing accuracy,

speed, and quality.
• The chosen model should be simple, practical, and useful.

19
 Step 5: Model Evaluation and Interpretation

The model is evaluated to identify the algorithm providing the best solution.

The algorithm is initially run on the validation dataset.

If the validation shows high accuracy, the model can be thought to predict
new cases and address the problem.

20
 Step 6: Model and Results Communication

It is key for the analyst to present the model in a way other people can
understand, particularly management.
• A good approach is to collaborate with key stakeholders early-on.
• A full understanding of the model is important.
• Whether simple or complex, the model should be explainable in
straightforward terms with appropriate visualizations.

21
 Step 7: Model Development

The model is not finished until it has been implemented and running on
real-time records to offer decisions or actions.
• This step involves other key stakeholders who need trained to
implement the system.
A key consideration throughout the 7-step marketing analytics modeling
process is to evaluate the ethical dimensions of the analysis.
• Are the privacy and anonymity of the subjects being protected?
• Does a bias exist in the data that could impact the analytics results?
• Are the model results accurate?
• The model may be correct, but the objective is unfair to some subjects
or unrealistic in its predictions.
• Another issue is that the data, features, data cleaning, and the model
are determined by analysts – ethics is imperative.

22
 Ethical Considerations

Advanced analytics has the potential to provide valuable insights into

customer behavior.
• But there is a significant risk they can be used to harm customers.
• Consumers are more concerned than ever about how companies are
collecting and using their personal data.
Companies must pay attention to ethical considerations of the use of AI
and machine learning in marketing applications.
• Be transparent about how data is collected and used.
• Obtain consent before collecting data.
• Ensure that algorithms do not contain bias or discriminate against
individuals.

23