Subscription Churn Prediction

End-to-end ML pipeline for predicting 30-day subscription churn and optimizing retention spend.

The problem

Subscription companies lose money when they offer blanket discounts to users who would have renewed anyway. The question is: can we predict who's actually at risk, and target our retention budget accordingly?

What this does

1. Predicts churn probability for each user (30-day window)
2. Segments users by lifetime value using K-Means
3. Simulates ROI of targeted vs. blanket retention campaigns

The goal is to spend retention dollars only where they matter.

Documentation

• Business Impact & ROI - Why this matters financially
• Data & Engineering - How the pipeline works
• Model Performance - Metrics and interpretation

Project structure

├── data/          # raw and processed data + saved model (gitignored)
├── docs/          # MkDocs site
├── figures/       # generated plots
├── reports/       # metrics.json model card (reproducible)
├── src/           # pipeline code
├── tests/         # pytest
├── Makefile
└── pyproject.toml