35% less churnData Science
Customer Churn Prediction
ML-powered churn prediction identifying at-risk customers 30 days before cancellation with explainable insights.
35%
Churn Reduction
30d
Early Warning
88%
Accuracy
$4.2M
Revenue Saved
The Challenge
5% churn rate class imbalance. CS team needed explainable predictions with recommended actions.
The Solution
1
SMOTE + calibrated probability scoring
2
40+ signal feature engineering
3
SHAP explanations in CS dashboard
4
Automated intervention triggers by risk tier
// Tech Stack
Built With
Python
XGBoost
Pandas
FastAPI
PostgreSQL
Airflow
SHAP
Streamlit
// Results
Impact
Churn dropped 35%, saving $4.2M ARR. CS team proactively engages 30 days before cancellation.
// More