Unit 1: Case Studies — Introduction to Data Mining

Case 1

Personalization at Scale

Netflix Recommendation Engine

Netflix uses data mining to analyze viewing patterns of 200+ million subscribers. By mining watch history, ratings, and browsing behavior, Netflix predicts what users want to watch next, saving $1 billion per year in customer retention.

Netflix collects 100+ million viewing events daily.
The recommendation algorithm accounts for 80% of content watched.
Netflix uses collaborative filtering and content-based mining.
A/B testing drives continuous algorithm improvement.

Focus Question:

What type of data mining does Netflix primarily use for recommendations? A. Classification | B. Association Rule Mining | C. Clustering | D. Regression

Tip: Netflix is a classic example of association rules (users who watched X also watched Y) combined with collaborative filtering.

Case 2

Market Basket Analysis

Walmart "Beer and Diapers" Discovery

Walmart discovered through data mining that men who buy diapers on Friday evenings also tend to buy beer. This unexpected association led to placing beer near diaper displays, boosting sales by 35%. This is one of the most famous association rule mining examples.

The discovery was made through association rule mining on transaction data.
Support and confidence metrics identified the pattern.
The correlation was behavioral, not causal.
Walmart uses this approach across 11,000+ stores globally.

Focus Question:

Which data mining technique discovered the beer-diapers connection? A. Classification | B. Clustering | C. Association Rule Mining | D. Anomaly Detection

Tip: Market basket analysis uses support, confidence, and lift to find unexpected product associations in retail data.

Case 3

Predictive Analytics in Marketing

Target Pregnancy Prediction

Target assigned each customer a "pregnancy prediction" score using data mining on purchase history. By analyzing changes in buying patterns (unscented lotion, supplements, cotton balls), Target could predict pregnancy and send targeted coupons — sometimes before families knew.

Target analyzed 25 product categories to build a pregnancy prediction model.
The model achieved 87% accuracy in predicting due dates.
Customer ID linked purchase history across visits.
The story raised major privacy and ethics concerns.

Focus Question:

What ethical concern did this case raise? A. Copyright infringement | B. Privacy violation | C. Data loss | D. Algorithm bias

Tip: Predictive analytics can be powerful but must balance business value with customer privacy and ethical considerations.