End-to-end data science project optimizing Starbucks offer targeting through customer segmentation (K-Means, k=4), predictive modeling (XGBoost AUC-ROC=0.994), and a rule-based recommendation system achieving +7.9% lift with rigorous causal inference and statistical validation.
How it knows: Segments are only reported after a stability check (ARI=0.85), and the causal claims carry bootstrap confidence intervals and propensity score matching rather than raw group differences.
Nicholas Smith - all work