An interactive D3.js dashboard analyzing revenue, churn, and customer segments for a telecom subscription business (7,043 real customers). Calculates "Revenue at Risk" turning a churn rate into a churn cost and finds that contract type is the strongest churn predictor (42.7% for month-to-month vs. 2.8% for two-year contracts), while payment method is a surprisingly strong signal on its own: electronic check users churn at 45.3%, nearly 3x automatic payment methods.

A Python data pipeline analyzing how BMW, Mercedes-Benz, and Audi depreciate compared to Ford, using 87,000 real Craigslist listings. Caught and fixed a major data quality issue (55% of raw data was duplicate listings), quantified value retention by brand, and trained a Random Forest model predicting price within $4,200 average error. Audi retains value best of the three luxury brands at every time horizon.