0% found this document useful (0 votes)

933 views165 pages

Marketing Analytics PDF Book

Uploaded by

Rushikesh Pandharpatte

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

933 views165 pages

Marketing Analytics PDF Book

Uploaded by

Rushikesh Pandharpatte

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 165

MARKETING ANALYTICS

PG Program

Subject Code- S3EL4

www.mitsde.com
MARKETING ANALYTICS - S3EL4

Chapter 1 - Introduction

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 Understand which marketing decisions are supported by
analytics.
 Learn what different analyses are important considering
product categories.
 Differentiate between business intelligence, analytics, and
data science.
 Learn about exploratory data analysis, predictive
analytics, and prescriptive analytics.
 Understand organization of the book and the use cases.

MARKETING ANALYTICS
 Marketing Analytics is a discipline that uses Statistics and
Machine Learning models and visualizations to provide
data-driven inference for marketing decisions.
 Market Research is the discipline that examines feasibility
of new product or service by using research conducted
 Business model itself determines data availability and
data gathering mechanisms
 Marketing decisions are made to a varying extent using
analytical tools on data

MIT School of Distance Education 2

MARKETING ANALYTICS - S3EL4

DATA FOR MARKETING ANALYTICS

 Data available with the firm may not be enough and not
always readily useful for inference
 We would think of business to consumer (B2C) firms to be
typical businesses with marketing analytics use cases
 B2B selling does use increasing data with various
channels
 For post-Internet era businesses he data are at the center
of marketing

WHAT ARE BUSINESS INTELLIGENCE,

ANALYTICS AND DATA SCIENCE
 For every marketing decision that firms want to make
basis data, it needs careful assessment of quality of data
 The market research firms take much care to gather
useful retail level samples to estimate reliable market
shares
 Data science is the science of collecting, representing,
and analyzing data using statistical, machine learning
techniques to derive useful actionable insights

Marketing Analytics(by Seema Gupta and

Marketing Analytics (by Seema Gupta
Avadhoot and Avadhoot
Jathar) CopyrightJathar) Copyright © 2021 by Wiley India Pvt. Ltd
2021 byWiley
India Pvt. Ltd

MIT School of Distance Education 3

MARKETING ANALYTICS - S3EL4

ANALYSIS
 done primarily to answer why questions
 usually retrospective, in purely explaining an event of the
past
 Analysis as a term is qualified with many domain specific
keywords

EXPLORATORY DATA ANALYSIS

 Forms the first key step in helping gauge the story with
the data.
 Business performance is measured by metrics widely
called Key Performance Indicators (KPIs).
 EDA typically involves many simple statistical techniques
 Product reviews are key for the way in which consumers
think about a brand, besides branding the firm may do

cont..

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

 We read the data with the following snippet of code.

 Textual data are to be represented as numeric, so

preprocessing of text is needed
 We usethe “spacy” package in Python for this

cont..

 In the next piece of code, we evaluate the sentiment for

each review.

 We would like to examine the sentiment across all the

reviews, and thus a histogram-like plot would help

MIT School of Distance Education 5

MARKETING ANALYTICS - S3EL4

 We derive a few characteristics, in addition to sentiment

 two metrics are derived as count of words and length of
reviews
 examine correlation with sentiment, as done by code and
represented by the plot of correlation

 It is easier to see that the sentiment is negatively correlated

with review length as well as word count
 long review is likely to be associated with not so higher a
sentiment

MIT School of Distance Education 6

MARKETING ANALYTICS - S3EL4

 one’s interest would be to understand what words or terms

occur often in reviews
 want to understand what do most reviews talk of

 A sample for consumer reviews about restaurant, they

write about Indian food, sweet, quality, coffee, breakfast,
dosa, and so on

MIT School of Distance Education 7

MARKETING ANALYTICS - S3EL4

DESCRIPTIVE ANALYSIS
 EDA but typically uses cross tabulation or statistical
summarization for inference.
 Descriptive analysis derives insights based on managers/
leaders’ understanding of the context of the metrics.
 For example, if the firm gets only a quarterly single data
point for a couple of years, about the brand share and,
therefore, not a good enough sample for statistical analysis
 firm has its own internal marketing spends data
 hence analysis could check directional evolution of these
two metrics

PREDICTIVE ANALYSIS
 Predictive analytics is a set of techniques to provide
estimate or a forecast of a metric for some futuristic
purposes.
 Predictive analysis answers the first question on what will
happen
 A company that successfully uses predictive use cases is
proactive in making informed decisions
 A commonplace marketing predictive use case is
determination of credit score.

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

PRESCRIPTIVE ANALYSIS
 Prescriptive analytics is a set of tools and techniques that
help assess different courses of action and usually
recommends best actions based on payoffs.
 explains which actions are useful and what consequences
of each action would achieve objectives with KPIs such as
sales, churn rate, or conversions
 It typically uses some form of artificial intelligence
techniques—statistical and machine learning models—
along with optimizations or simulations.

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 2 - Segmentation

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics(by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 Understand the types and benefits of customer analytics.
 Understand how data is collected and analyzed in
segmentation.
 Explore the process, applications, and examples of cluster
analysis.
 Perform calculations using clustering algorithms such as
k-means.
 Interpret and evaluate clusters.
 Understand hybrid segmentation and other tools used for
clustering.

Consumer Needs and Segmentation

Analysis

MIT School of Distance Education 2

MARKETING ANALYTICS - S3EL4

 the company decided to form mini-campaigns to address

each segment’s needs.
 mini-campaign would utilize various media platforms to
communicate the marketing messages carefully to each of
the segments.
 Customer Analytics It is a field that uses customer-level
data – primarily behavioural data along with demographics
to understand, and predict customer behaviour.
 Customer analytics plays an important role in the rediction
of customer behaviour
 Customer analytics uses various techniques such as
predictive modeling, information management,
egmentation, and data visualization to predict consumer
behavior
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

Benefits of Customer Analytics

 Using customer analytics, you can find out which
marketing channels work the best for your company.
 It helps in identifying which customers are on the verge of
leaving your company
 If you have decoded customer purchase patterns, that
means you have won half the battle because it is the key
to increasing sales in your business, and customer
analytics provides you the pathway to do so.
 Customer analytics answers questions on how to
differentiate between such customers and highlights ways
to focus on more profitable customers

MIT School of Distance Education 3

MARKETING ANALYTICS - S3EL4

Factors Essential for Obtaining Benefits

from Customer Analytics
 A survey reports that more than 85% of the firms who use
customer analytics extensively, report enormous growth in
their company.
 When these reports were compared with low-level (20%) and
mid-level (30%) customer analytics users.
 it is essential to gain the utmost excellence in customer
analytics to obtain the most benefit.
 It is observed that only 28% reported a significant growth in
the company where the senior-level executives were not
involved, whereas 69% reported considerable growth where
senior-level executives were involved

Segmentation Analytics
Data Collection in Segmentation:
 a great way to stay connected with customers and also to
fetch details about them.
 Second, try fetching details from customer services.
 Third, can gather information about them from digital
platforms such as social media.
 Fourth, method is through online behavior tracking by
using cookies

Marketing Analytics(by Seema Gupta and

Marketing Analytics (by Seema Gupta
Avadhoot and Avadhoot
Jathar) CopyrightJathar) Copyright © 2021 by Wiley India Pvt. Ltd
2021 byWiley
India Pvt. Ltd

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

Customer Segmentation Process

 Define the purpose of segmentation

 Identify the most crucial variable
 Value each segment
 Improve your products

Data Analysis for Customer Segmentation

After you are done collecting data, you need to analyze it to
gain insights. For this, you can follow the following
procedure

 Data tabulation
 Cross-tabbing
 Cluster analysis
 Size estimation of each segment
 Valuation of each segment

MIT School of Distance Education 5

MARKETING ANALYTICS - S3EL4

Cluster Analysis
 Clustering is a method to discover groupings in a dataset.
 The purpose of clustering is to simplify a large dataset
into meaningful subgroups.
 we can use statistical techniques such as clustering, based
on which, we can easily define the similarities and
differences
 for segmenting customers, firms, or markets, we can see
that creating subgroups in a dataset is inherently
subjective
 purpose of clustering is to divide the customer into
subgroups allow marketers to differentiate their approach
by segments in order to maximize customer value
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

Application of Cluster Analysis

 Distance measures: Distance measures are actually a

measure of dissimilarity, with larger values denoting lesser
similarity

where dij is the Euclidean distance between the data points i

and j
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

Interpreting Clusters
There are two aspects to interpreting clusters:-
 What cluster members have in common: The centroid is
used to define a typical member of the cluster

 How each cluster is different from other clusters:

 This is key to differentiate segments for marketers to
evolve a customized approach to market to each segment

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

Evaluating Clusters
 Cluster diameter: maximum distance between any two
points within the cluster and indicates the maximum
dissimilarity between members of the same cluster.
 Cluster variance: sum of the squared distance from the
centroid of the cluster.
 Cluster silhouette: The silhouette score of a clustering
solution is the average of the silhouette scores of all
individual customers (data points) in the customer base

Assigning New Customers to Clusters

 Compute the distance between the new customer and
each cluster centroid.
 Assign the new customer to the closest cluster
 it is prudent to define a threshold for cluster diameter so
that the maximum distance between the new customer
and existing customers.
 In attitudinal segmentation, input data may be collected
using a primary survey for clustering.
 predictive models are used to predict clusters based on
variables

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

Hybrid Segmentation
 It is possible to develop multiple segmentation solutions
 Each segmentation uses a different lens
 Micro segmentation is often useful for personalizing
communication using marketing technology

K-Means Clustering: Example

we have the following information: (a) gender, (b) age, (c) in
come, and (d) siblings.
Setting Up the Excel Sheet:

MIT School of Distance Education 11

 Copy the formulae from L13:R13 to L14:R268 to complete

computations for all Ids
 In Q9, compute the sum of squared distances for all Ids
from their closest cluster center SUM(Q13:Q268)

Using Solver to Arrive at Optimal Cluster Centers

 Enter Solver parameters as shown to find the optimal
cluster centers

MIT School of Distance Education 14

MARKETING ANALYTICS - S3EL4

 Run Solver to find the optimal cluster centers and

assignment of each ID to the closest cluster

Interpretation of Clusters
 U13:V17 give the count and percent of customers belonging
to each cluster
 W13:W17 give the variance in each cluster
 X13:Z17 provide the average values for attributes within
each cluster
 AA13 to AA17 provide the percentage of Females in the
cluster
 X18:AA18 have the same values for the entire dataset

MIT School of Distance Education 15

MARKETING ANALYTICS - S3EL4

 Cluster 1: Higher income, fewer siblings, average age,

average percentage of females
 Cluster 2: Much younger (minors), very few siblings, very
low income
 Cluster 3: Many more siblings, average on all other
attributes
 Cluster 4: Much older, few siblings, average income, higher
percentage of females
 Cluster 5: Relatively older, few siblings, 6X income, and
average gender distribution.

Number of Clusters

 Till now, we used Excel, same method can be used to

develop two-, three-, or four-cluster solutions as well
 The elbow chart shows how the variance drops substantially
from 2 to 3 to 4 cluster solutions and starts stabilizing at 5

MIT School of Distance Education 16

MARKETING ANALYTICS - S3EL4

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 3 - Positioning

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics(by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

that uses the following
determinant attributes:
 To what extent does the
product appear to have
lower or higher amounts
of sugar.
 To what extent does the
product appear to have
lower or higher amounts
of caffeine.

Using Multiple Product Attributes

 Correspondence analysis A statistical technique used to
showcase interrelated data as plot points in a two
dimensional setting

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

Joint Perceptual Maps

 show the required
needs of various
market segments based
on the same attributes.
 The word “joint” is
used because the
segment requirements
are mapped hand in
hand with the
perceived positioning
of the product

Constructing a Perceptual Map

 choose two determinant attributes

 The second step is to make a list of competitors that are

to be plotted onto a perceptual map

MIT School of Distance Education 5

MARKETING ANALYTICS - S3EL4

 The ratings are fictitious survey ratings for the two

attributes of trendiness and durability for the five clothing
retailers A, B, C, D, and E.

 Based on this map, we can infer that brand E has a

negative impression on its consumer

White Spaces
Case Study: How McDonalds Used White Spaces to Maximize
Profits
 started out as a tiny drive-in eatery serving barbeque and
burgers to customers in cars
 McDonalds saw this as an opportunity to reinvent their
business model
 Through an extensive study of their consumer base, they
found that most of their sales came solely from two items:
burgers and potato chips.
 They reduced the menu such that it only contained burgers
and potato chips, and scrapped the drive-in
 facility to make a walk-up window instead

MIT School of Distance Education 6

MARKETING ANALYTICS - S3EL4

And
Avadhoot Jathar

© U Copyright
Marketing Analytics(by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 Know about the choice models.
 Learn conjoint analysis: principle, variables, steps, benefits,
and software.
 Understand the market forecasting and the S curve.
 Know about the Bass diffusion model and adopter
categories.

Introduction
 The qualitative analysis of data to understand how users
interact with a product.
 The product can be related to consumer goods or even
IT products such as a banking app or a food-ordering app
 The actual users’ interaction with the product provides
rich understanding of usage patterns and preferences and
helps companies to correct the product features or
redesign better products

MIT School of Distance Education 2

MARKETING ANALYTICS - S3EL4

Analyzing Digital Products

 the case of digital products, the data are available at the
level of each customer.
 to discern what is working for many customers or for a
product team to implement analytics using sophisticated
techniques that could be done once a large customer
base is available

an exhausting task for an individual respondent.
 Sample choicetasks would be like:
Which of the following products would you choose?
• HP, Intel@ Celeron N4020 Processor, 6 GB RAM, 1 TB HDD,
14 inch black @ `36,190
• Lenovo, AMD Ryzen-3 Processor, 4 GB, 256 GB SSD, 14 inch
Platinum Grey @` 43,249
• Mi Notebook, Intel Corei7 Processor, 4 GB RAM, 256 GB SSD,
Black 15.6 inches @` 43,249

 Respondents provide their preferred choice of products for

such choice tasks

Market Forecasting
 When you have no history of the sales of products,
prediction of future product sales must be done with
market-level understanding of how new products or
innovations are adopted by consumers and, therefore, we
turn to diffusion theory.
 for forecasting new product sales: S curves and Bass
Diffusion Model.

MIT School of Distance Education 12

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 5 – Pricing

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 List and explain the goals of pricing.
 Understand pricing strategies such as bundling, skimming,
and revenue management.
 Elucidate the relevance of pricing strategies to
products/service categories.
 Explain the concept of price promotions.
 List the various types of price promotions and discounts.
 Enumerate the ways to evaluate which price promotions
work.
 Comprehend price elasticity modeling and using
regression analysis for the same.

Introduction
 It is the approach taken by a firm to set the price for the
product or services it sells.
 A company may be guided by a pricing objective it sets for
itself
 pricing may define further levers available to shape the
demand by considering decisions such as price increases,
price discounts, or price coordination

MIT School of Distance Education 2

To sum up, for skimming pricing strategy to work:

 The product must have novelty
 competitors may not be able to imitate thefeatures easily
 price-inelastic consumers who are willing to buy the
product at premium prices
 work with products with unique features that will justify
initial premium pricing
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

MIT School of Distance Education 7

MARKETING ANALYTICS - S3EL4

Analytics with Price Skimming

 price point can be decided based on understanding the
willingness to pay of the relative price-inelastic consumer
group for the product.
 judging the willingness to pay of consumers would
require an elaborate conjoint analysis exercise.
 downward price revisions must be judged on the basis of
the elasticity of the product category

Revenue Management
 sell a product at a lower price today in order to sell it at a
higher price tomorrow.
 The main challenge is to determine the demand of the
product over time.
 there are time dimensions to each sale
 Both revenue management in airlines and markdown
pricing in fashion make use of prices to cater to different
consumer segments while still achieving better revenue.

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

Promotions
 Refers to setting price lower than actual list price for a
short term to attract price-sensitive customers and boost
sales by selling to them
 Some promotional offers come in the form of loyalty or
cashback points
 increasing loyalty-oriented (encouraging repeat purchase)
promotions are popular with retailers

Measuring Promotional Lifts

 simple approach to lift measurement is not quite correct
as there could be several other factors
 For this formula to be useful, the non-promotional and
promotional timeframes must be reasonably short

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

Types of Promotions
 Special event pricing
 Cash rebates
 Low-interest financing
 Warranties and service contrasts
 Psychological discounting
 Loss leader pricing:

Discounting
 This refers to lower prices to consumers than the list
prices, and can manifest in several ways such as discount
coupons, cashbacks, or quality discounts or seasonal sales
and so on.

Types of Discounting
 Functional (trade) discounting
 Quantity discounting
 Cash discounting
 Seasonal discounting
 Geographical discounts

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

Price Elasticity of a Beverage Brand

 Price elasticity measures change in demand due to a unit
change in price.
 percent change interpretation that how many % change
will be achieved in volume (quantity) for 1% change in
price.

 A leading whiskey marketer is interested in understanding

the elasticity of its premier whiskey brand.
 Whiskey is a nonessential product that falls in the
category of indulgence consumption with consumers
seasoned and loyal to the taste

 Before attempting a model with price elasticity, it is essential

to explore data. The brand has three SKU sizes—180 ml, 350
ml, and 700 ml.

 We wish to evaluate the extent to which these increases

eroded volume sales, if at all.

MIT School of Distance Education 11

MARKETING ANALYTICS - S3EL4

Figure shows trade promotion spends by brand with the

wholesalers. This is an occasional spend with no specific
time identified consistently over time by the brand.

Transformations on Metric in Price

Elasticity Model
 Let us examine this closely as a regression model,
comprising only price and intercept variable:

Marketing Analytics
Chapter 6 – Marketing Mix

Dr. Seema Gupta

And
Avadhoot Jathar

 Base Variables: not influenced by marketing activities but
are dependent upon various other factors such as brand
equity, GDP, growth rate, and consumer sentiment
 Distribution variables: number of stores where the product
is available, the shelf life of the product , and several stock-
keeping units
 Seasonality: variables used for capturing time and season-
specific sales patterns
 Macro-economic variables:not directly under the control of
a company but they do affect them substantially
 Industry-level variables: specific factors outside the control
of a firm/brand
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

INCREMENTAL VARIABLES
 Above-the-line (ATL) marketing: marketing activities that
are non-targeted and possess a wider reach.

 Below-the-line (BTL) marketing: marketing activities that

are targeted towards a particular audience and are very
specific and directive in nature

MIT School of Distance Education 4

ESSENTIAL EVALUATION METRICS IN

REGRESSION
 Fit measures are useful for the relative comparison of
models with 5 variables and 8 variables.
 market mix modelling with 8 variables and better fit.

Adjusted R Squared

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

 Variance Inflation Factor

Here, Rx1 2 is the R squared (explained variance) value

obtained by regressing one covariate x1 on the rest of the
(p −1) covariates [x2 ,x3 ,…,xp ] in the model.

HETEROSKEDASTICITY
 assumption that regression error variance does not
increase or decrease with systematic increase in values of
explanatory variables.

 An important assumption regarding the regression error

is that it has constant variance and it is conditionally
independent of the explanatory variables in X.

 The estimate βˆ on log transformed ad spends actually

has an elasticity interpretation, that is, percent change in
ad-spend would result in βˆ percent change in sales.

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

Predictive/Forecasting Ability Metrics

such as MAE/MAPE
 Mean Absolute Error (MAE) measures the average of the
absolute deviation of predicted values from actual values

 Other Considerations in Relation to Data

there are also a few other issues that a modeler needs to be
conscious of while building regression models,

 Missing Data: Missing data needs to be handled before

using a metric in a model.
 Outliers: It is an observation that has a completely different
realized value of a metric as compared to the rest of the
sample data.

 Scaling Variables: important from the point of view of

understanding the relative importance of variables through
coefficient estimates βˆ using regression

 Variable Interactions: By specifying each of the media

MARKETING ANALYTICS - S3EL4

APPLYING PRINCIPAL
COMPONENTS TO BRAND
EVALUATIONS DATA
 The R program for applying principal components.
 The program can be found as .R file
‘ch3_loyal_cust_engagement_use_case.R’.
 Library ‘readxl’ helps read the excel sheet and ‘factoextra’ helps with
visualization of principal components.

APPLYING PRINCIPAL
COMPONENTS TO BRAND
EVALUATIONS DATA
 We then set the working directory (folder on your drive) where input files are
placed and read the input file using ‘read_excel’ command.

 We then drop the column in the data frame that has the customer serial number.
 This is the first column and the following code accomplishes this task.

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

APPLYING PRINCIPAL
COMPONENTS TO BRAND
EVALUATIONS DATA
 Scree plot plots eigenvalues of principal components with each principal
component in the model. The eigenvalues of components are in descending order.
 The eigenvalues represent variance of each successive principal component and
thus scree plot is used to select few components to represent enough variation in
the data.

Scree plot.

APPLYING PRINCIPAL
COMPONENTS TO BRAND
EVALUATIONS DATA
 Another popular interpretation of principal components is to be examined by data
visualization.
 The original metric and the principal components can be represented on a
scatter plot as:

MIT School of Distance Education 11

MARKETING ANALYTICS - S3EL4

APPLYING PRINCIPAL
COMPONENTS TO BRAND
EVALUATIONS DATA
 PCA is used in non-survey datasets for dimension reduction. For
example, images are high-dimensional data, and principal
components can be used to represent the data with only a few
variables (components).
 Also, principal components is a method to represent a large
number of variables with a smaller set of transformed variables
called components.
 These components are usually used in inference, which may be t-
statistics with the same component, or using components as
predictor variables in regression or supervised classification
techniques.

MIT School of Distance Education 12

MARKETING ANALYTICS - S3EL4

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 8 – Nurturing Customers

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 Define customer experience analytics and understand its
goals.
 Understand the different metrics used for tracking
customer experience such as net promoter score,
customer effort score, churn rate, and upsell and cross-
sell rates.
 Understand logistic regression analysis and how to use
diagnostic tests and use logistic regression analysis as one
of the classification techniques.

INTRODUCTION
 Customer experience: A holistic view of how consumer
interacts with the brand/product during all stages of
product purchase
 Consideration
 Purchase as well as post-purchase engagements in the
form of customer’s cognitive, psychological, and
behavioral responses.

MIT School of Distance Education 2

MARKETING ANALYTICS - S3EL4

METRICS FOR TRACKING

CUSTOMER EXPERIENCE
 Net Promoter Score (NPS)
 Customer Effort Score (CES)
 Churn Rate
 Upsell and Cross-sell rate

Customer Feedback Metrics

 Typically obtained from small survey responses that

firms obtain from their customers.
 Net Promoter Score: A single customer-centric metric that
determines the overall customer opinion/satisfaction about your
product/brand/service.
 Based on the scores, customers can be classified into one of the
following three types:
1. Promoters, whose rating is 9–10;
2. Passives, whose rating is 7–8;
3. Detractors, whose rating is 0–6.

Net promoter score.

NPS = % Promoters – % Detractors

MIT School of Distance Education 3

MARKETING ANALYTICS - S3EL4

Customer Feedback Metrics

 Customer Satisfaction Score: Metric used to measure
how satisfied a customer is with the use of a
product/service.

Smileys for satisfaction.

 Customer Effort Score: It measures how difficult or

easy it was for a customer to get the work done.
 Usage of CES and NPS

Behavior-Derived Customer Metric

 Churn Rate: Percentage of customer base that is lost

over a specific length of time, due to customers
switching to competitor products or switching out of
purchases in the category.
 Upsell and Cross-Sell Rates:
 Upselling: To recommend or sell a premium priced product
similar to customer’s current preference but a product with
better value or higher quality.
 Cross-selling: To recommend or sell products that
complement the original product.

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

CLASSIFICATION TECHNIQUE
From the R program for logistic regression, the following code reads the
data:

The next task is to split the data into a training set and a test set.

USE OF LOGISTIC REGRESSION AS A

CLASSIFICATION TECHNIQUE
A customer with lower total annual spends will have all other spends lower as
well. For the first iteration, we omit the annual spend and half-yearly spend
metrics to keep the issue of multicollinearity in check.

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

USE OF LOGISTIC REGRESSION AS A

CLASSIFICATION TECHNIQUE
To examine model performance in the test sample, let us predict the
response
based on test data metric with the created logit model object.

We compare the predictions with actuals in a confusion matrix, also called a

PART OF BUYER PERSONA

 Sales representatives
 Marketing representatives
 Executive leadership
 Past customers

CUSTOMER LIFETIME VALUE

 It is a measure to determine how much a customer is
going to spend on your business over the whole period
of your relationship, or it can be defined as a metric to
calculate the total future profit you can expect from a
customer.
 The main advantage of focusing more on CLV is that it
enables you to make an efficient strategy to target the
required customers because all customers are not the
same.

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

NEED FOR CALCULATING CLV

 A number of parameters are involved in calculating
how many customers you acquired in a day or week.
 These parameters include click-through rate, bounce
rates, conversion rates, etc.
 Simply knowing that you are getting a lot of new
customers is not enough.
 Retaining old customers is as important as getting new
ones.

CALCULATING CLV

 The simplest one requires only three components:

 Average order value (AVG)
 Purchase frequency (PF)
 Customer lifetime length (CFL)

MIT School of Distance Education 5

MARKETING ANALYTICS - S3EL4

BOOSTING CLV
 You can create a forum to actively answer any doubts
of customers. Form relationships with the customers,
and both parties can learn from each other.
 You should be able to analyze the trends and be up to
date with the customer’s journey. By utilizing these
data, you can figure out what improvements can be
made for the betterment of the customer’s journey.
 Only updating customers about a particular product is
not enough. You should be able to send customized
and personalized messages to the customers to satisfy
their needs. The key here is to engage with the
customers intelligently.

IDENTIFYING PROFITABLE
CUSTOMERS
 Historical CLV
 It is found by identifying the past behavior of customers
without thinking about their future behavior.
 Predictive CLV
 It holds information about the past behaviors and future
expected retention of the customers to estimate their
lifetime value.

MIT School of Distance Education 6

MARKETING ANALYTICS - S3EL4

COMPUTING CLV

where r is the probability of the customer

buying/being alive at time t, T is the time horizon

for estimating the CLV, i is the customer index, t
is the time index, and d is discount rate.
The three key parameters in CLV calculation:
Contribution margin
Retention rate
Time horizon
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

EXAMPLE OF CLV
 Simple CLV

MIT School of Distance Education 7

MARKETING ANALYTICS - S3EL4

EXAMPLE OF CLV
 CLV Computation with Annual Forecasts and
Discounting

EXAMPLE OF CLV
 Adding Uncertainty to Parameters in Computing CLV

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

CHURN ANALYTICS

 It is the analysis of the customer loss of a firm.

 Examples of how a customer leaves a company:
 1. Unsubscribing from updates.
 2. Deleting their accounts.
 3. Choosing another company.
 4. Nonrenewal of contract

CALCULATING CHURN RATE

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

IDENTIFYING CHURN

 Subscription churn: It means when there is a certain

period for which the customer and the company are in
a relationship. After the period is over, it is up to the
customer to decide whether to renew the subscription
or leave.
 Non-subscription: It is up to the customer to end the
relationship with the company at any time.

IMPORTANCE OF CUSTOMER RETENTION

 It has proven records of driving higher profits

 According to a report by McKinsey, the executive
teams that actively use customer data analytics in their
business observe an increase of 126% in profit than
the ones who do not use analytics.

Probability of selling to existing and

new customers.13

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 10 – Digital Analytics: Metrics
and Measurement

Dr. Seema Gupta

And
Avadhoot Jathar

information about the
places that were clicked
maximum number of
times on your webpage.

 On-site surveys:
companies typically ask
questions such as:
(a) What elements were you
looking for but could not
find?
(b) What barriers were
stopping you?

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

How do you reduce exit rates?

 Slow loading speed of the page

 Issues in navigation
 Substandard UI/UX
 Distracting content

Bounce Rate
 Bounce rate This gives information about the proportion
of visitors who leave your website after going through
just one webpage
 Some of the reasons for visitors leaving may be:
1. The site takes too long to load a page.
2. An error page may be loaded.
3. Users are unable to find what they are looking for.
4. Users find what they are looking for but are not
compelled enough to proceedfurther.

users did not find what they were looking
left your site without adding anything to their cart

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

Unique Visitors
 The individual people who visit the site over a specific
period of time.
 Even if a visitor visits the same site multiple times in a
given period, it will still be counted as one visit.

Attribution Challenge and Shapley

Regression
 it is important to understand that conversions must be
attributed rightly to the various clicks of a customer
before an eventual purchase.
 One of the ways of correct attribution is to use the game
theory approach known as Shapley value analysis

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

 Shapley values can be derived using the following

equation

 The following code defines all possible subsets of channel

combinations that a customer may have been exposed to
before making a purchase

Marketing Analytics(by Seema Gupta and

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

 function accumulates the value of each coalition of

channels, that is, each combination of channels, and is the
value function of the set ѵ(S) described

 calculate the value functions of combinations by using the

function written earlier

MIT School of Distance Education 11

 Assuming unequal sized groups, the total within-group

variance can be written as:

 is, the deviance of grand mean from each group mean,

multiplied by the number of observations in each group.

Marketing Analytics(by Seema Gupta and

 The ratio of variances as a test statistic is distributed as an

F-distribution
 The degrees of freedom for the F-distribution are (n – m,m
– 2), where n is the total observations from m groups

MIT School of Distance Education 14

MARKETING ANALYTICS - S3EL4

ANOVA: Linear Regression Model

 The regression specification is as follows:

 In the preceding regression equation, yij is the stacked

data of all groups j = {1, 2,…,m}, ԑij is the econometric
error, and i denotes every observation
 a null hypothesis to test as
 the F-statistic as it is defined as

Search Engine Marketing

 is a digital marketing strategy consisting of website
promotion for increasing visibility on the SERP through
paid efforts

 Why Is SEM Important?

1. an essential part of online marketing to expand a
company’s reach.
2. targets visitors at accurate times when they require new
information.
3. results are immediate

MIT School of Distance Education 15

MARKETING ANALYTICS - S3EL4

SEM Platforms
 The most famous ad platform is Google Ads. There are
two networks offered by Google

1. Google Search Network

2. Google Display Network

Marketing Analytics(by Seema Gupta and

How Does an Ad Win the Ad Auction

 it is not necessary that every single ad appears on the
search results page
 Quality Score is a metric based on an ad’s overall quality,
and maximum bid is the amount an advertiser is ready to
pay for a click on an ad

MIT School of Distance Education 16

MARKETING ANALYTICS - S3EL4

Search Engine Optimization

 Search engine optimization is the procedure of expanding
traffic on site with the help of organic search results on a
search engine

 Working of SEO
1. Search engine crawling
2. Search engine indexing
3. Search engine ranking

Factors Affecting SEO

 Some factors that influence on-page SEO are
1. Title tag
2. Meta description
3. ALT tag
4. Internal links
5. Sub-headings

Marketing Analytics(by Seema Gupta and

MIT School of Distance Education 17

MARKETING ANALYTICS - S3EL4

Off-Page SEO
There are multiple ways in which you can use off-page SEO
to your benefit:
 Links
 Trust
 Social

Marketing Analytics(by Seema Gupta and

Social Media Analytics

 The following are some of the most important metrics
that a business needs to track
 Social share of voice (SSoV):

 Social media conversion rate:

 Post reach

MIT School of Distance Education 18

MARKETING ANALYTICS - S3EL4

App Marketing Metrics

 If you have a mobile app, then the following metrics are
important
 Retention
 Engagement
Session level, Session interval, Opt-ins, Opt-outs,
Interactions
 Cost per install (CPI)
 Average revenue per user (ARPU)

MIT School of Distance Education 19

MARKETING ANALYTICS - S3EL4

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 11 -Artificial Intelligence and
Machine Learning

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 Understand tenets of deep learning (DL) and the
components of an artificial intelligence (AI) system.
 Think of and understand marketing use cases of AI/ML
algorithms in direct marketing, personalization, and
recommender systems.
 Understand thoughts/concepts involved with using text,
images data for marketing applications.
 Understand in-depth implementation details two
important use cases – targeting (direct marketing) and
recommendations – understand use of Python libraries
for such purposes.
 Understand the challenges involved with AI.

INTRODUCTION
 Artificial intelligence (AI) is a simulation of human intelligence by
machines and its applications typical span natural language
processing, processing image/video data.
 Machine learning is a branch of Artificial Intelligence where
computer algorithms can perform certain tasks and improve this
performance through experience and by the use of data.
 Deep learning is a branch of Machine Learning and uses neural
network algorithms for the learning task. Deep learning is used with
data, such as images and text requiring natural language
processing.
 Recommender systems are systems designed to recommend
things (content, products, or services) to users. The
recommendations are based on user’s past behavior (things user
engages with) or behavior of users similar to given users.

MIT School of Distance Education 2

MARKETING ANALYTICS - S3EL4

IMPORTANCE OF AI IN MARKETING
 It is extremely essential for businesses to understand
how AI can reshape consumer behavior and how to
exploit the same for business objectives.
 Primary objectives of AI in marketing can be:
 Identifying right target audience
 Branding
 Growth and financial improvements
 Chatbots is a software application that does online chat via text or text-to-
speech in lieu of direct contact with human agent.
 Content curation

KEY APPLICATIONS OF AI IN MARKETING

Personalization of Online Experience
 Personalization is done in the form of couponing and recommendations for
products, or even content that an individual may choose to engage with.
 Personalization is done with the website or app content shown to an
individual customer.

Benefits of Personalization – Marketing Professionals survey.

MIT School of Distance Education 3

MARKETING ANALYTICS - S3EL4

KEY APPLICATIONS OF AI IN MARKETING

Chatbots
 Chatbots use AI-backed algorithms to offer online customer support and
guidance.
 Chatbots typically handle commonly recurring customer service issues very
well.

Kik’s customized options.

KEY APPLICATIONS OF AI IN MARKETING

AI-Powered Dynamic Emailing
 This is a form of personalization of emails done
for each individual customer.
 The following are essential tasks of dynamic
emailing:
 Predictive send-time optimization
 Predictive subject lines
 Dynamically managing email assets such as templates

MIT School of Distance Education 4

MARKETING ANALYTICS - S3EL4

COMMON TERMINOLOGIES – AI, ML,

AND DL
 AI as a discipline points to the potential of machines to imitate human like
intelligence in one way or another.
 ML algorithms refer to algorithms using repetitive sampling techniques for
achieving prediction and pattern recognition tasks with data.
 DL is a subset of ML, which in turn is a subset of AI.
 DL is a branch that uses multiple layers of neural networks for pattern
recognition and prediction tasks.
 Neural networks are computing systems inspired by the biological neural
networks that constitute animal brains. It is based on a collection of
connected units or nodes called artificial neurons, which loosely model the
neurons in a biological brain.
 ML models are key subset of methods of AI, we examine key concepts of ML
models.

COMMON TERMINOLOGIES – AI, ML,

AND DL
Relationship between AI, ML, and Deep Learning.

MIT School of Distance Education 5

MARKETING ANALYTICS - S3EL4

IMPORTANT CONCEPTS OF ML
 Techniques such as regression and logistic
regression are traditionally understood and
applied as single sample inference.
 The predictive ability of these techniques can
suffer severely if new data is different from the
single sample data with which the regression
models are built.
 Resampling techniques, thus, involve splitting
data and developing many models to help
achieve better predictive ability in the new
dataset that was not used to build the models.

VARIANCE AND BIAS TRADE OFF

 The difference between the true value and the predicted value is
called the error.
 The error called as bias is the error arising due to wrong
assumptions about the learning algorithm.
 The bias would arise because of assuming a linear relationship
among variables than a relationship that may be non-linear.
 Training data is used to build the model.
 To avoid overfitting, we examine the concept of variance of the
model. We do not want the variance of the model to be too large,
that is, overfitted model that performs poorly in the data that the
model has not yet seen.

MIT School of Distance Education 6

MARKETING ANALYTICS - S3EL4

TRAINING
 Process of building a model where an ML or DL algorithm
understands the complex relationship between labelled data and
predictor features.
 The process of building models typically entails some objective
minimization of distance between the actual label data and the
label data predicted by a model.
 Data annotation is a process of creating training dataset.
 For an image classification problem, images are labeled with a
class and this class becomes a categorical target variable.
 The predictive performance of model parameters is checked on
the validation dataset.
 There are additional parameters with resampling methods, known
as hyperparameters, that help with achieving least error metric in
the validation dataset.

VALIDATION AND TEST

 Predictive accuracy is the key to any ML modelling task.
 Model validation includes assessing how good the model is in
predicting labels correctly in new data, which the model has not
yet seen.
 Three subsets of data are usually deployed in the process of
training, validation, and testing.
 A holdout dataset, called the test dataset is where the
performance of the model (prediction error) is checked.
 This idea is extended to k-fold cross-validation.
 Random forests is an ensemble learning method that combines
the outcomes of several decision trees built independently with
randomized feature space in each tree.

MIT School of Distance Education 7

MARKETING ANALYTICS - S3EL4

THE k-FOLD CROSS-VALIDATION

 Cross-validation uses repeated sampling techniques.
 Other ways to decide the value of k are as follows:
 Representative
 Leave one out cross-validation

Splitting the data for k-fold cross-validation.

THE k-FOLD CROSS-VALIDATION

Averaging with k-fold validation.

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

BOOTSTRAPPING/BAGGING VALIDATION
 The training dataset is chosen at random.
 A key difference between bootstrapping and cross-validation is
that in the former, the average error rate changes in every
iteration.
 Bagging is used as a concept in random forest model, combining
inference of several tree-based models.

Bagging/bootstrapped averaging.

REGULARIZATION PENALTY
 Regularization penalty methods avoid overfitting models.
 The penalty methods change the objective function to include a
penalty term.
 Regularization penalties are either based on the addition of the
square of model parameters to form a modified objective function
(ridge regression with L2 penalty), or the addition of the modulus
of model parameters to form a modified objective function
(LASSO regression with L1 penalty).
 Targeting has traditionally been done through demographics-
based rules and past engagement with a customer at a very basic
level.
 Booking a fixed deposit or not is a binary label.
 The problem for targeting is a classification problem.

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

DECISION TREE
 A decision support tool that uses tree-like model of decisions and
the possible outcomes. Decision trees are partitioning algorithms
and are supervised techniques.
 The individual data with incomes, monthly expenses, and fixed
deposit booking is represented in a two-dimensional space.

Two-dimensional space and partitions.

RANDOM FORESTS
 The modification from a simple bagging model.
 Random forest algorithm builds several decorrelated
trees and averages predictions.
 Extending this to random forests where multiple trees are
built, following are the parameters that decide accuracy:
 1. Number of trees grown
 2. Number of features (variables) used at random in
building a tree
 3. Maximum depth of any tree
 4. Minimum number of samples to split a node in a tree
 5. Criteria with which to split a node (entropy or Gini
index for classification and Mean squared error for
regression).
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

Data Description and Formatting before

Classification Model
 We import numpy and pandas libraries for handling data and read
the dataset with read_csv of pandas.
 The data is read by the following two lines of Python code:

Data Description and Formatting before

Classification Model
 We check the variable y from the dataset as the label indicating whether a
customer booked a deposit or not. It is coded as yes or no, it must be
converted into a 1/0 variable. This is done by user-defined function, here
is_yes(), that takes a string argument and returns 1 if string has ‘yes’ or
otherwise it returns 0.

MIT School of Distance Education 11

MARKETING ANALYTICS - S3EL4

Data Description and Formatting before

Classification Model
 We then check the data types of the variables in dataframe. This is done
using code: bank_marketing.dtypes

Data Description and Formatting before

Classification Model
 The indexes for these columns are given as a list in iloc.

We use describe() function with dataframe bank_marketing.

MARKETING ANALYTICS - S3EL4

BOOSTING TREES
 Boosting is an important concept in ML.
 The boosting classifier can be thought of as sequentially applying
classification algorithms on repeatedly modified versions of data.

ROC for gradient boosting classifier on test dataset

VARIABLE IMPORTANCE
 A tree is very good for visualization (if it is not too deep)
and for explaining splits as coded rules, random forests
with multiple trees do not have the advantage of such
simplicity.
 Across the trees, any variables that consistently help to
improve the split are ranked in the variable importance
chart.
 The variable Euribor 3-month rate and number related to
employment are the key variables in determining whether
a customer books a fixed deposit.
 Neural networks also work on the philosophy of
understanding a highly non-linear mapping of inputs to
outputs.
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

MIT School of Distance Education 15

MARKETING ANALYTICS - S3EL4

SIMPLE FEED-FORWARD NETWORK

 Each node in every layer is connected to all the other nodes in
the previous layer.
 The connection is non-cyclic in nature and information moves
unidirectionally from the input layer to the hidden layer to the
output layer in this type of network.
 Activation functions: Mathematical transformations to connect
nodes in layers of a neural network – activation functions provide
way to map inputs to an output and choice of them depends on
nature of outcome variable and difficulties with optimizing weights
of a neural network.

Simple Neural Network

with 2 inputs, 1 output, 1
hidden layer with 2
nodes.

DEEP NEURAL NETWORK

 A deep neural network contains numerous (N) hidden layers of
nodes between the input and output layers.
 Information flows through the ‘N’ number of layers and the
probability of the outcome is computed at each layer.
 Deep neural networks have many applications, such as
mathematical modelling, speech to text recognition, and computer
architecture.

Diagrammatic
representation of deep
neural network.

MIT School of Distance Education 16

MARKETING ANALYTICS - S3EL4

IMAGE RECOGNITION
 Ubiquitous in security, transportation – driverless car technology,
defense systems, and ones that we regularly come across are
identification of people for security checks, captcha on websites
for allowing access to genuine users (as opposed to bots), and
improved experience in augmented reality gaming.
 Image recognition consists typically of one of the following three
tasks:
 Image classification
 Single object localization
 Object detection
 Based on the distinct features, it must identify class of feature:
 Collection and categorization of data
 Creating image features
 Classification–Learning algorithm

CONVOLUTIONAL NEURAL NETWORK

 A convolutional layer is used to analyze a part of an image.

Image and filter with 3 × 3 window.

Process of convolution
and feature map.

MIT School of Distance Education 17

MARKETING ANALYTICS - S3EL4

CONVOLUTIONAL NEURAL NETWORK

Scan of an image with Prototype CNN architecture for image

zero padding, with a filter classification – four Convolution
(3  3). layers with Max pooling.

MARKETING APPLICATIONS
OF IMAGE RECOGNITION

 Visual listening in
 Search experience coupled with images
 Auto tagging
 Recommendations based on interests
 Faster product checkouts
 Product information with an image

MIT School of Distance Education 18

MARKETING ANALYTICS - S3EL4

WORKING WITH TEXTUAL DATA

 Textual data is commonplace for marketers
with product reviews, responses on social
media to campaigns or on twitter reflecting
sentiment about the products or services.
 The central themes are:
 Representing text data
 Model

RECURRENT NEURAL NETWORK

 Recurrent Neural Network (RNN), the network is
designed to take in a series of inputs, that is, not
just multiple values but a series that has
meaning in its temporal sequence.
 RNNs, therefore, can take inputs with no fixed
size vector (as simple neural networks would do)
but do need to remember the sequencing in the
inputs.
 RNNs have many applications in sentiment
analysis, autocorrection, and autosuggestion of
text; all tasks are within the domain of natural
language processing.
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Copyright © 2021 by Wiley India Pvt. Ltd

MIT School of Distance Education 19

MARKETING ANALYTICS - S3EL4

RECOMMENDATION SYSTEMS
 Popularity-based recommendation
 Content-based recommendation

Following are two primary ways recommender systems are designed:

1. Collaborative filtering
2. Content-based filtering

COLLABORATIVE FILTERING
 Collaborative filtering is a method of making automatic predictions
(filtering) about the interests of a user by collecting preferences or
taste information from many users (collaborating).
 There are mainly the following two steps in this process:
 Search for consumers who have viewing patterns that are similar to that
of a sample user.
 The patterns obtained can be used to compute a new prediction for the
sample user.

Collaborative
filtering.

MIT School of Distance Education 20

MARKETING ANALYTICS - S3EL4

CHALLENGES INVOLVED WITH AI

 Ethical and privacy issues
 Loss of human capital
 Lack of high-quality data and prohibitive infrastructure
costs

Challenges
associated with AI in
marketing.

MIT School of Distance Education 21

MARKETING ANALYTICS - S3EL4

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

Marketing Analytics
Chapter 12 - Data Visualization

Dr. Seema Gupta

And
Avadhoot Jathar

© U Copyright
Marketing Analytics (by Seema Gupta and Avadhoot Jathar) Dinesh Kumar,
© 2021 IIM Bangalore
by Wiley India Pvt. Ltd

MIT School of Distance Education 1

MARKETING ANALYTICS - S3EL4

LEARNING OBJECTIVES
 Understand why visualization matters.
 Create important visualization charts with univariate,
multivariate data, and descriptive visualizations.
 Learn key visualization charts with some analytics
techniques.

MARKETING ANALYTICS - S3EL4

BAR CHART

Vertical bar chart: over time

comparison of same metric,
sales.

Horizontal bar chart: sales in

geographical markets at the
same point in time.

STACKED BAR GRAPH

 Can compare the whole data category parts.

Stacked bar chart to compare data across quarter

and year (multiple cuts).

MIT School of Distance Education 7

MARKETING ANALYTICS - S3EL4

HISTOGRAM
 bar graph used when grouped data are represented.

Histogram: bar chart with intervals.

BOX PLOT
 A method used to represent grouped data in their
quartiles.
 Also employed to understand outliers if any in a variable.

Box plot: distribution

of variable over time

MIT School of Distance Education 8

MARKETING ANALYTICS - S3EL4

VIOLIN PLOT
 small variation on the box plot where the kernel density
plot is superimposed on each side of the box plot.
 The density plot is also a way of understanding how a
random variable is distributed.

PIE CHART
 Circular graphical representation charts of different categories for which a metric
sums up to 100%.
 Each section of the pie chart represents the proportion of the category in the
whole.
 Variation of pie chart, called doughnut chart, is extremely useful for providing the
percentage share of one category in the total market as opposed to many different
categories represented in a pie chart.

Pie chart: share of the whole, by each

entity.

MIT School of Distance Education 9

MARKETING ANALYTICS - S3EL4

SCATTER PLOT
 Used to examine the relationship between any two
metrics.
 Useful for examining the relationship between two
continuous variables as opposed to continuous and
categorical variables.
 An extension of scatter plots is the scatter plot matrix.

Relationship between sales

and marketing spend.

BUBBLE CHART
 Represent data with three dimensions.
 It is an extension of scatter plot.

Bubble chart: plotting a

third variable values as
size of dot on scatter plot.

MIT School of Distance Education 10

MARKETING ANALYTICS - S3EL4

LINE CHART
 Formed simply by systematically connecting various
points plotted in the graph.
 Useful to represent data in a timeline.

Line chart for representing variation over time.

HEAT MAP
 Used to visualize the density of a specific data using color
dimensions.
 Use a warm-to-cool color spectrum to show variation in values
by gradation.

Heat map.

MIT School of Distance Education 11

MARKETING ANALYTICS - S3EL4

VISUALIZATIONS

VISUALIZATIONS USEFUL WITH

COMMON DATA SCIENCE TECHNIQUES
 Visualizations are not confined to just variables or metric alone
in isolation.
 Central also to communicating inference from
statistical/machine learning techniques.
 Any basic predictive models with sales requires two aspects:
 Visualizing sales or brand’s volume share or profit like measure for a
certain marketing input.
 (a) This could be done with a bar chart or a pie chart.
 Ability to change the marketing inputs and see the incremental change
in sales/share or profits.
 (a) Changing the marketing input requires an interactive editable input cells on
user interface, which can allow users to examine input values and output bar/pie
chart in the same visualization.

MIT School of Distance Education 12

MARKETING ANALYTICS - S3EL4

VISUALIZATIONS FOR COMMUNICATING

REGRESSION INFERENCE

 Regression analysis provides insights that are important in two ways:

prediction for given input and marginal effect of each input.
 Analytics practitioners think of representing regression inference in three
ways for marketing analytics use cases:
 Decomposition charts
 Elasticity and decomposition chart

Decomposition chart: year-on-year change.

VISUALIZATIONS FOR COMMUNICATING

REGRESSION INFERENCE

Trend charts with regression

Trend chart with sales regression and important predictor price.

MIT School of Distance Education 13

MARKETING ANALYTICS - S3EL4

Visualizations for Communicating

Regression Inference
Important variables using bar plot/box plot.
 Red and blue colors denote variables with negative and positive significant
effects, respectively.

Regression coefficients and confidence intervals

VISUALIZATION INFERENCE FOR

PREDICTIVE MACHINE LEARNING
MODELS
 With machine learning techniques (gradient boosted trees, random forests, neural
networks) that are predictive techniques with repeated sampling done primarily to
improve cross-validation predictability, one of the key graphs is variable importance
chart.

Variable importance.

MIT School of Distance Education 14

MARKETING ANALYTICS - S3EL4

COMMUNICATING INFERENCE WITH

SEGMENTATION/CLUSTERING
 Predominantly two clustering methods find applications with marketing
analytics:
 K-means (or K-median) clustering used typically to segment consumers.
 Hierarchical clustering to segment markets (or identify markets that are more alike).

Visualizing a cluster
solution.

COMMUNICATING INFERENCE WITH

SEGMENTATION/CLUSTERING
 Segmentation refers to the process of categorizing groups of data based on
their similarities.

Hierarchical
clustering: market
segmentation

MIT School of Distance Education 15

MARKETING ANALYTICS - S3EL4

CONCLUSION
 Visualization is key to communicating a story as seen with
the data.
 In this objective, the multiple chart dashboards are
common with marketing practitioners.
 We discussed key charts and their uses for visualizing
typical marketing data.
 We have also considered how model-based inference is
communicated to business leaders using simple
visualization charts.

MIT School of Distance Education 16

Modern Synthetic Methods - 4
No ratings yet
Modern Synthetic Methods - 4
49 pages
Gastrointestinal Physiology
No ratings yet
Gastrointestinal Physiology
6 pages
CM19352 Process Optimization
No ratings yet
CM19352 Process Optimization
2 pages
Matching Ending
No ratings yet
Matching Ending
3 pages
Tata Car Insurance 2024-25
No ratings yet
Tata Car Insurance 2024-25
8 pages
Mga Batayan Sa Pagpili NG Asignaturang Medyor NG Mga Magaaral NG Batsilyer NG Sekundaryang Edukasyon Sa Pampamahalaang Unibersidad NG Kanlurang Mindanao
No ratings yet
Mga Batayan Sa Pagpili NG Asignaturang Medyor NG Mga Magaaral NG Batsilyer NG Sekundaryang Edukasyon Sa Pampamahalaang Unibersidad NG Kanlurang Mindanao
14 pages
Vistas Prose 3 - Journey To The End of The Earth
No ratings yet
Vistas Prose 3 - Journey To The End of The Earth
4 pages
APAT06 Day 2
No ratings yet
APAT06 Day 2
19 pages
Eijppr 2023 Vol 13 Iss2 April 1 17 2116
No ratings yet
Eijppr 2023 Vol 13 Iss2 April 1 17 2116
17 pages
Iso 25539-4-2021
100% (1)
Iso 25539-4-2021
18 pages
Design and Installation of Deep Benchmarks
No ratings yet
Design and Installation of Deep Benchmarks
8 pages
Simulation of Passive Earth Pressure Against Retaining Wall Considering Wall Movement Mode
No ratings yet
Simulation of Passive Earth Pressure Against Retaining Wall Considering Wall Movement Mode
9 pages
Principles: Life and Work
From Everand
Principles: Life and Work
Ray Dalio
4/5 (648)
Telephone Call 1
No ratings yet
Telephone Call 1
36 pages
Rubrics For Student Engagement or Class Participation
No ratings yet
Rubrics For Student Engagement or Class Participation
2 pages
The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are
From Everand
The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are
Brené Brown
4/5 (1175)
Freeze Concentration of Coffe and Tea
No ratings yet
Freeze Concentration of Coffe and Tea
1 page
Physical Exam-1 2022 Revamp
No ratings yet
Physical Exam-1 2022 Revamp
6 pages
RHUB5921 Description
No ratings yet
RHUB5921 Description
11 pages
Film Poster Proposal 1
No ratings yet
Film Poster Proposal 1
3 pages
The Glass Castle: A Memoir
From Everand
The Glass Castle: A Memoir
Jeannette Walls
4.5/5 (1856)
The Emperor of All Maladies: A Biography of Cancer
From Everand
The Emperor of All Maladies: A Biography of Cancer
Siddhartha Mukherjee
4.5/5 (298)
Sing, Unburied, Sing: A Novel
From Everand
Sing, Unburied, Sing: A Novel
Jesmyn Ward
4/5 (1267)
Gonzales Gene08 Ethics Course Syllabus
No ratings yet
Gonzales Gene08 Ethics Course Syllabus
7 pages
Steve Jobs
From Everand
Steve Jobs
Walter Isaacson
4.5/5 (1139)
Shoe Dog: A Memoir by the Creator of Nike
From Everand
Shoe Dog: A Memoir by the Creator of Nike
Phil Knight
4.5/5 (629)
The Outsider: A Novel
From Everand
The Outsider: A Novel
Stephen King
4/5 (2886)
Her Body and Other Parties: Stories
From Everand
Her Body and Other Parties: Stories
Carmen Maria Machado
4/5 (903)
Enjoying English Grammar
No ratings yet
Enjoying English Grammar
232 pages
Byproducts of Sulfur Hexafluoride (SF) Use in The Electric Power Industry
No ratings yet
Byproducts of Sulfur Hexafluoride (SF) Use in The Electric Power Industry
11 pages
The Perks of Being a Wallflower
From Everand
The Perks of Being a Wallflower
Stephen Chbosky
4.5/5 (4103)
Angela's Ashes: A Memoir
From Everand
Angela's Ashes: A Memoir
Frank McCourt
4.5/5 (943)
Fear: Trump in the White House
From Everand
Fear: Trump in the White House
Bob Woodward
3.5/5 (836)
The World Is Flat 3.0: A Brief History of the Twenty-first Century
From Everand
The World Is Flat 3.0: A Brief History of the Twenty-first Century
Thomas L. Friedman
3.5/5 (2289)
Rise of ISIS: A Threat We Can't Ignore
From Everand
Rise of ISIS: A Threat We Can't Ignore
Jay Sekulow
3.5/5 (144)
A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story
From Everand
A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story
Dave Eggers
3.5/5 (233)
Team of Rivals: The Political Genius of Abraham Lincoln
From Everand
Team of Rivals: The Political Genius of Abraham Lincoln
Doris Kearns Goodwin
4.5/5 (244)
2 Disbursement Request Form
No ratings yet
2 Disbursement Request Form
1 page
The Yellow House: A Memoir (2019 National Book Award Winner)
From Everand
The Yellow House: A Memoir (2019 National Book Award Winner)
Sarah M. Broom
4/5 (100)
John Adams
From Everand
John Adams
David McCullough
4.5/5 (2546)
Guieline Full
No ratings yet
Guieline Full
460 pages
Ir 4426
No ratings yet
Ir 4426
12 pages
Manhattan Beach: A Novel
From Everand
Manhattan Beach: A Novel
Jennifer Egan
3.5/5 (919)
Little Women
From Everand
Little Women
Louisa May Alcott
4.5/5 (2369)
The Light Between Oceans: A Novel
From Everand
The Light Between Oceans: A Novel
M.L. Stedman
4.5/5 (815)
50 Year Perspective of Automotive 50 Year Perspective of Automotive Engineering Body Materials and An Engineering Body Materials and An Analysis of The Future Analysis of The Future
No ratings yet
50 Year Perspective of Automotive 50 Year Perspective of Automotive Engineering Body Materials and An Engineering Body Materials and An Analysis of The Future Analysis of The Future
33 pages
Research 7 Q3 W4
No ratings yet
Research 7 Q3 W4
9 pages
A History of The Cavendish Laboratory 1871-1910 PDF
No ratings yet
A History of The Cavendish Laboratory 1871-1910 PDF
388 pages
Usace Eng Form 4025-r
No ratings yet
Usace Eng Form 4025-r
2 pages
The Unwinding: An Inner History of the New America
From Everand
The Unwinding: An Inner History of the New America
George Packer
4/5 (45)
Trust Receipts Law
No ratings yet
Trust Receipts Law
5 pages
SC Form 2 Chapter 1
No ratings yet
SC Form 2 Chapter 1
37 pages