Overview
What this challenge is about.
Use 18 months of anonymized data (provided) covering: usage events, login frequency, support tickets, NPS responses, billing health, feature adoption, practice firmographics. Define churn label (cancelled within 90 days), build train/validation/test splits with time-aware cutoffs (no leakage). Train baseline (logistic regression) + a gradient-boosted model (XGBoost or LightGBM). Evaluate: precision-at-top-200, AUC, calibration plot. Run SHAP analysis to surface top drivers per at-risk account. Deliver: model code + tests, 8-page evaluation report, top-200 watchlist for next month, and a 4-page CS playbook on what to do per driver pattern.
The Brief
What you'll do, and what you'll demonstrate.
Build a churn-prediction model that beats CS gut-feel on precision-at-top-200 and is paired with an actionable playbook.
Earning criteria — what you'll demonstrate
- Apply supervised learning to a real churn-prediction task with proper splits
- Compare baseline vs gradient-boosted models on business-relevant metrics
- Use SHAP to translate model behavior into intervention guidance
- Pair ML output with operational playbook for measurable impact
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Product Manager
PMs on retention products need this end-to-end ML-to-playbook fluency to scope features that actually move net-revenue retention.
This challenge sharpens
- business-analytics
- ml-applications
- model-evaluation