Marketing & Retail Analytics

Introduction

@ Confidential Documents – Prepared by Rajarshi Pandit

What it takes to become Analytics Practioner?

Techniques

Stat – Math-ML • Statistical Modeling

• Machine Learning

• R, Python
• SAS
Business Skills
• Tableau, Power BI etc. Tools/Technology
• Domain Knowledge
Domain Knowledge
Coding • Interpretation of
algorithms

Problem Solving Skill - Analytical Ability – Data/Number Crunching – Willingness to learn

@ Confidential Documents – Prepared by Rajarshi Pandit 2

 What are the key components of Analytics?
Analytics Value Chain

• Strategic analysis
4 Predictive & Prescriptive
• Heavy use of statistics, machine learning & optimization
Analytics techniques

• Deep dive analysis to answer the “W” questions

3 • Combination of business knowledge, data query
Diagnostic Analytics
techniques & basic statistics

• Develop and maintain business Reports & dashboard

2 Descriptive Analytics • Data query & reporting using standard tools

Data Engineering / • Structured & Unstructured data

1 • Extract, Transform, Load data
Management
• Data integration and aggregation

Value Chain Activities

@ Confidential Documents – Prepared by Rajarshi Pandit
 Use Case: Customer Retention

Business Problem Statement: Retailer is spending lot of money to acquire customers, but customers are not staying
with the business. Current churn rate is around 25%

Data sources Data Engineering & storage BI Dashboards Analytics objectives Business objectives

Customer details
Data Cubes/Tables Understand the
customers who are
Reports / Dashboards churning out
- Tableau, PowerBI etc (Profile Customers )
Customer behavior/ Retain
transaction details 1 2,3 Analyze data to develop Customers
customer persona
Adhoc Diagnostic Analytics
Data Storage: Predict future Churn
Customer service / On-premise (SQL (Develop Churn
Customer complain data Server, Oracle etc.) Prediction Model)
Cloud (AWS, Azure
etc)

•Data acquisition & integration

•Use of tools like SQL / Python •Descriptive & Diagnostic Analytics
•Reading & Converting textual data • Tools: BI tools / Excel
using NLP •Methods: Hypothesis testing, Data partitioning, etc Example: Descriptive Analytics and visualization
@ Confidential Documents – Prepared by Rajarshi Pandit
 Use Case: Customer Retention

Business Problem Statement: Retailer is spending lot of money to acquire customers, but customers are not staying
with the business. Current churn rate is around 25%

Data sources Data Engineering & storage ML / Statistics /Optimization Analytics objectives Business objectives

Customer details
Analytics Data Mart Understand the
customers who are
churning out
(Profile Customers ) Retain
Customer Transaction
details 1 4 Customers

Data Storage: Predict future Churn

On-premise (SQL Modeling & Optimization
Customer service / (Develop Churn
Customer complain data Server, Oracle etc.) Prediction Model)
Cloud (AWS, Azure
etc)

Probability of
•Data acquisition & integration
Churn in future
•Use of tools like SQL / Python •Predictive & Prescriptive Analytics
•Reading & Converting textual data • Tools: Python, R, SAS etc Recommendation
using NLP • Methods: Logistic Regression, Decision for Retention
@ Confidential Documents – Prepared by Rajarshi Pandit Tree, A/B Testing Initiatives
 How to create a solution framework to address the business problem?
• Convert the business problem statement into analytics problem statement
• Explain each step what you are going to do and how that is helping to solve the problem
• Mention data requirements, analytics methods and expected outcome from each stage of the process

Solution framework (example)

Testing hypothesis & Predict future behavior Provide actionable
Use relevant data
understanding customers of the customers recommendations

Data Preparation: Collection, validation Exploratory Analysis / Data Identify opportunities &
Predict future churn
, aggregation and integration Visualization recommend retention initiatives

Gather relevant data Exploratory Analysis Model development Insights & Recommendation
• Customer details / Demographics • Testing relation between customer • Test multiple classification • Use exploratory analysis findings,
• Customer transactions details demographics and churn behavior techniques like Logistic regression, model results and churn probability
• Product details • Testing transaction behavior with decision tree to develop churn scores to identify opportunities for
• Customer service / customer complain customer churn prediction model retention initiatives
data • Profile customers to understand • Use model with best prediction • Estimate expected business impact
Feature Creation who is churning ability • Provide actionable
• Define churn indicator • Identify drivers for churn • Identify major factors driving recommendations
• Create behavior variable based on customer churn • Monitor impact of interventions
transaction Diagnostic Analysis • Use final model to estimate through dashboards
Aggregate and integrate • Use data partitioning methods like customer level churn probability
• Create customer level data by aggregating CHAID/CART to identify pattern score
transactions
• Integrate multiple data element to create
360-degree view of customers

@ Confidential Documents – Prepared by Rajarshi Pandit 6

 Other Use Cases

Use Case: Pricing Analytics

What will happen to sales if price is increased by 1 %?

You need to establish a relationship between Price and sales

Sales = f (Price, Marketing activities, promotions, Distributions, etc.)

Application of Statistical Regression – Multiple linear regression

Y = β0 +β1* Price + β2X2+…….+βkXk+ε

Y → Dependent Variable (Sales), X i→ Other Independent variables / Explanatory Variables (marketing,
Promotion, Distribution etc.)

@ Confidential Documents – Prepared by Rajarshi Pandit

 Other Use Cases

Use Case: Campaign Response Model

What is the probability that a customer will respond to a direct to customer

campaign?
You need to predict chance of occurrence of an event.

Application of Statistical Regression – Logistic regression / Classification Algorithms

𝑝
Log(1−𝑝) = 𝑎 + 𝑏 𝑥

P ➔ probability of customer response

@ Confidential Documents – Prepared by Rajarshi Pandit

 What method to use?

Pricing use case

• Dependent variable: Continuous variable (Sales)

Sales = β0 +β1 * Price

Linear Regression

Predicting Customer response

• Dependent variable: Binary Variable (Y = 0, 1), possible outcome

of an event
𝑒 (𝑎+𝑏𝑥) 𝑝
P= ➔ Log( ) =𝑎+𝑏𝑥
1+𝑒 (𝑎+𝑏𝑥) 1−𝑝

Classification techniques (e.g. Logistic regression)

@ Confidential Documents – Prepared by Rajarshi Pandit
 Introduction to Marketing Analytics

Customer Purchase Funnel Marketing

Above The Line Marketing

(ATL)

Below The Line Marketing

(BTL)

Direct to Customer
Marketing (DTC)
@ Confidential Documents – Prepared by Rajarshi Pandit 10
 Introduction to Marketing Analytics Channels Analytics Solutions

• TV • Marketing Mix
• Radio • Media Mix
• Print • Attribution
Media • Out of Home (OOH) • Channel Optimization
• A/B Testing
• Digital • etc
✓Search, Video, Banner,
Social, Influencer
Marketing, etc
• Promotion Effectiveness
• Coupons • Pricing Analytics
• Display • Test Control Analysis
Promotions • Feature / Circular • Sales Forecasting
• Assortment Analysis
• Multi Buy
• Combo Pack • Segmentation
• Propensity Modeling
• DTC • Loyalty Analytics
Direct to Customer Customer ✓ Email • Campaign Effectiveness
Marketing (DTC) Marketing ✓ SMS/WA • CLTV
✓ DM • Cross-Sell/ Up-Sell
@ Confidential Documents – Prepared by Rajarshi Pandit
• Churn Analysis 11
 Objectives of Marketing Analytics
• Improve Marketing Effectiveness:
• Improve Awareness→ Drive Consideration → Impact conversion/Sales

• Understand what marketing channels perform better and optimize marketing ROI by
allocating available marketing spend across channels

• Understand effectiveness of different promotions tactics and enable brand/category

managers to define better promotion plan to drive sales

• Maximize customer value by enabling business to take right decision at each stage of
customer life cycle (Acquire -> Grow -> Retain -> Win-back)
✓ Understanding Customer Preference
✓ Identify high valued customer
✓ Retain customer at risk
✓ Improve campaign effectiveness
@ Confidential Documents – Prepared by Rajarshi Pandit 12
 Basic Terminology in Retail

1. What is a SKU ?
– Stock Keeping Unit - SKU' A store's or catalog's product and service identification code, often
portrayed as a machine-readable bar code that helps the item to be tracked for inventory. These
codes are not regulated nor standardized. When a company receives items from a vendor, it has a
choice of maintaining the vendor's SKU or creating its own.

2. What is a UPC ?
– The Universal Product Code (UPC) is a barcode symbology (i.e., a specific type of barcode) that is
widely used in the United States, Canada, the United Kingdom, Australia, New Zealand, and in other
countries for tracking items in stores. Its most common form, the UPC-A, consists of 12 numerical
digits, which are uniquely assigned to each item

@ Confidential Documents – Prepared by Rajarshi Pandit

 Basic Terminology in Retail

3. What does CPG stand for? What does FMCG stand for?
– Consumer packaged goods (CPG) are consumable goods such as food and beverages, footwear and apparel, tobacco,
and cleaning products. In general, CPGs are things that get used up and have to be replaced frequently, in contrast to
items that people usually keep for a long time, such as cars and furniture. In India and in some other countries CPG
companies are called as FMCG companies, which stands for Fast Moving Consumer Goods.

4. What are the 3 most important attributes(variables)that we look at when

trying to understand customer behavior ?
1. What does a consumer buy
2. How often does he buy and when does he buy what he buys
3. How much does he spend

@ Confidential Documents – Prepared by Rajarshi Pandit

 Data Overview

@ Confidential Documents – Prepared by Rajarshi Pandit 15

 Data Overview: Product

Manufacturers

Manufacturer Cadbury / A
Mondelez Company
Category
Chocolates
Brand
Cadbury
Sub-Brand
Dairy Milk
Size
Small / 200g
UPC / EAN
12 / 13 digit code
PRODUCT
PRODUCTS HIERARCHY

@ Confidential Documents – Prepared by Rajarshi Pandit 16

 Measuring Sales

• Unit Sales
• Volume Sales
• Amount (dollar, rupees etc.) Sales

The neighborhood convenience store sold 65 300ml Coca Cola can last
week. Price per can is Rs. 20

• Unit sales last week: 65

• Amount Sales last week: 65 * 20 = Rs. 1300
• Total volume sales last week: 65 * 300 = 19,500ml

@ Confidential Documents – Prepared by Rajarshi Pandit 17

 Volume Equivalency

1 Unit = 1 Litre / 1000 ml 1 Unit = 200gms

1 Unit = 20 sticks 1 Unit = 66 Diapers

1 Unit = 1 Vol

@ Confidential Documents – Prepared by Rajarshi Pandit 18

 Data Measures

Why do we need Volumetric measures?

• Unit sales and volume sales might lead to different conclusion.

• While analyzing sales of a particular brand, there might be different pack sizes. Market
share measured in Units may be different from market share measured in volume.

Even though both brands have the same

unit sales (1,800) and unit share (50.0),
Brand B has higher volume sales and
volume share because of its bigger
package sizes. This is why it’s important
to look at volume sales – it leads you to
different conclusions.

[Link]
@ Confidential Documents – Prepared by Rajarshi Pandit 19
 Retail Data Dimensions

Transaction Data

Store Time Customer Product Measure

✓ Store No ✓ Year ✓ Customer Demo ✓ Category ✓ Units
✓ Store Location ✓ Month ✓ Customer ✓ Brand ✓ Values
✓ Store Segment ✓ Week Loyalty Data ✓ Sub-brand ✓ Price
✓ Day ✓ SKU / EAN

Marketing Data

Media ✓ Marketing Measurement (like TV GRP, Digital Impression, etc), Marketing Spend

Customer
✓ Past campaign details, Campaign Calendar, Campaign Response, Redemption data
Marketing

Promotion
✓ Promotion Calendar, Promotion Tactics, Promotional Spend
Data
@ Confidential Documents – Prepared by Rajarshi Pandit 22
 CPG/FMCG Data Dimensions

GEOGRAPHY TIME PRODUCT MEASURES

✓ National ✓ Year ✓ Category ✓ Volumetric
✓ Region ✓ Month ✓ Brand ✓ Price
✓ Market ✓ Week ✓ Sub-brand ✓ Promotional
✓ DMA ✓ Day ✓ PPG ✓ Distribution
✓ Store ✓ SKU
✓ ...
Marketing Data
Media ✓ Marketing Measurement (like TV GRP, Print TRP, Digital Impression,
etc), Marketing Spend

Direct to Customer ✓ Coupons, Direct Mails, etc

Trade Promotion
✓ Promotion Calendar, Promotional Spend, Price Discounts, etc

Other Promotion ✓ Sponsorship, FSI Coupons, Sampling, Shelf Talker, etc

@ Confidential Documents – Prepared by Rajarshi Pandit 23

 Example of Descriptive analysis and visualization

Distribution of Churn
Churn Trend
30.0%

CHURN RATE
26.50% 20.0%

73.50% 10.0%

0.0%
1 2 3 4 5 6 7 8 9 10 11 12
MONTHS
Churn

Distribution of Churn by Gender

Distribution by no of visits
55.0%
45.0%
45% 35.0% 36.0%

55%
20.0%

64% 9.0%
36% Back

Back
Churn Active 1-2 2-4 4+
Female Male Churn Active

@ Confidential Documents – Prepared by Rajarshi Pandit

 Example of Diagnostic analysis

Understanding customers who are Churn: 50

Non-Churn: 100 Churn Rate: 33.33%
churning out
Phone Technology
New Old

Churn: 30 Churn: 20
Churn Rate: 23.1%% Churn Rate: 100%
Non-Churn: 100 Non-Churn: 0

Month on Book
<= 27 Months > 27 Months

Churn: 25 Churn: 5
Churn Rate: 55.6%% Non-Churn: 20 Churn Rate: 0.06%
Non-Churn: 80

Age

<= 35 > 35

Churn: 20 Churn: 5 Churn Rate: 33.33%

Churn Rate: 100%
Non-Churn: 0 Non-Churn: 10

Back

Back
• Customers with Old technology will churn out
• Customers with New Tech, less than 27months on book, and les than 35 years old
@ Confidential Documents – Prepared by Rajarshi Pandit will churn out 29