Starbucks Capstone Challenge
Goal: Analyze customer behavior and predict responses to marketing offers.
Tech Stack: Python, Scikit-learn, Pandas, Jupyter, Supervised & Unsupervised Learning
Overview
Analyzed data from Starbucks’ mobile app to classify customers and predict which offers (BOGO, discount) they are most likely to respond to.
Approach
- Merged offer, transaction, and demographic datasets
- Engineered features for behavior, demographics, and recency
- Applied Random Forests and KMeans clustering
- Evaluated using ROC-AUC and adjusted accuracy
Results
- Segmented customers into 4 behavior clusters
- Improved offer targeting by 18% over baseline
- Provided strategic insights for personalized marketing
👉 View on GitHub