As part of BCG’s Data Science Virtual Program, I led a churn analysis for PowerCo, a major utility company.

Using Python tools like Pandas, scikit-learn, and Seaborn, I built a Random Forest model that improved the F1-score for churners from 0.10 to 0.30.

I engineered features from customer tenure, usage, pricing, and behavior, and addressed class imbalance with business-aware evaluation. Insights revealed price sensitivity and contract structure as key churn drivers.

The project simulated real consulting workflows, blending EDA, feature engineering, model tuning, and strategic recommendations.

👉 Check out the full notebook here

💻Check out my full Datacamp portfolio to see more projects and courses I’ve completed using Python, SQL, and more!