Data Science for Business (CS2988)

Case Study

Dr. Rourab Paul

Computer Science Department, SNU University, Chennai

​

​

Data Science For Business

Walmart & Hurricane Frances 2004

​

Data Science for Business

2

Walmart & Hurricane Frances 2004

• They will not focus heavily on mathematical or technical details
• Instead, they will show how data science helps in understanding real-world business problems

​

Situation

• Hurricane Frances was expected to hit Florida.
• People were evacuating and preparing for the storm.
• Walmart executives (in Arkansas, far away) saw this as a chance to use data-driven prediction.

​

Data Science for Business

3

What Walmart Wanted to Do

Instead of:

• Waiting for people to buy items
• Reacting after shelves were empty�

Walmart wanted to:

Predict what customers would buy before the hurricane arrived

This is the key idea:� Move from reactive decision-making to predictive decision-making

​

Data Science for Business

4

Why Simple Intuition Is Not Enough

Some things are obvious:

• People will buy bottled water
• People will buy flashlights

These do not require data science.

But:

• How much water?
• Which stores?
• Which specific products?
• Are these patterns local or nationwide?

These questions cannot be answered by intuition alone.

​

Data Science for Business

5

Role of Data Mining

Walmart analysts:

• Looked at past hurricane data (e.g., Hurricane Charley)
• Used huge historical sales data from their data warehouse
• Compared sales before hurricanes vs normal times
• Looked for unusual local patterns, not general trends

Surprising Discovery (Key Insight)

They found something unexpected:

• Strawberry Pop-Tarts sold 7× more than usual
• Beer was the top-selling pre-hurricane item

These are non-obvious patterns:

• No one would intuitively expect Pop-Tarts to spike
• Without data mining, this insight would be missed

​

​

Data Science for Business

6

Why This Is Important

​

Because of this insight, Walmart could:

• Stock the right products before the hurricane
• Avoid empty shelves
• Increase sales
• Improve customer satisfaction�

This is predictive analytics in action.

Why do strawberry Pop-Tarts sell 7× more before a hurricane?

Data Science for Business

7

Why This Is Important

​

“Data—not common sense—found the pattern.”

​

Data science discovers non-obvious patterns that change decisions.

​

Data Science for Business

8

Business Scenario: MegaTelCo

Data Science for Business

9

Business Scenario: MegaTelCo

• MegaTelCo is a large telecom company
• In one region, 20% of customers leave when their contracts expire
• The market is saturated → getting new customers is hard and expensive
• Companies now compete mainly by stealing customers from each other
• This switching behavior is called churn.

Data Science for Business

10

Why Churn Is a Serious Problem

• Losing customers means lost revenue�

​

​

Data Science for Business

11

• Attracting new customers requires:
• Discounts
• Promotions
• Marketing expenses�
• Retaining customers is cheaper than acquiring new ones

Therefore, reducing churn is a top business priority

Role as a Data Scientist

• Help understand why customers leave�
• Decide which customers should receive a retention offer�
• Use data to maximize the impact of a limited marketing budget

​

Data Science for Business

12

Important Detail: Budget Constraint

MegaTelCo:

• Cannot offer incentives to all customers
• Has a limited incentive budget
• Needs to choose customers strategically�

So the key question becomes:

Which customers should be offered the retention deal to reduce churn the most?

​

Data Science for Business

13

Why This Is Not a Simple Problem

It may seem easy to:

• Offer the deal to all customers
• Or offer it randomly�

But this is inefficient because:

• Some customers would stay without any offer
• Some customers will leave even with an offer
• Offers should be given only to customers most likely to churn and most likely to be influenced

Data Science for Business

14

Role of Data Science

Data science helps by:

• Analyzing historical customer data
• Identifying patterns that predict churn
• Estimating churn probability for each customer
• Ranking customers by risk of churn
• Selecting customers that provide the highest return on incentive spend
• This is predictive modeling.

Data Science for Business

15

Data-driven decision-making (DDD)

Data Science for Business

16

Evidence: Data-Driven Decisions Work

Data Science for Business

17

What the research shows

• Economists Erik Brynjolfsson and colleagues (MIT & Wharton) studied many firms
• They created a DDD score that measures:�� How strongly a company uses data across its decision-making processes�

Key findings

• Firms that are more data-driven are more productive
• This result holds even after accounting for other factors (size, industry, technology, etc.)
• The effect is large and statistically significant:

A one-standard-deviation increase in DDD → 4–6% increase in productivity

Data Processing Technologies

Data Science for Business

18

Big Data: Data that is too large or complex for traditional data processing systems.

Big Data = Warehouse full of raw materials

Data Science = Factory that turns materials into useful products

Core Functions of Big Data Technology

Data Science for Business

19

Big data technologies mainly support data engineering, not decision-making directly.

They are used for:

• Data storage
• Data ingestion
• Data processing (batch & real-time)
• Data management
• Supporting analytics and data science

Thank You

Data Science for Business

20