AI/ML & Data
Customer Churn Prediction
Ranking customers by churn risk — AUC 0.91, honestly reported.
A churn model that ranks customers by leave-risk so retention effort lands where it matters — reported with full metrics, not just the flattering ones.
01
The problem
Retention teams treat every customer alike because nobody knows who's about to leave — so saves go to the loud, not the leaving.
02
Our approach
Feature engineering over behavior and account signals, model comparison, and threshold tuning for the retention workflow it feeds.
03
The outcome
Strong ranking power (AUC 0.91) with balanced precision-recall (F1 0.69) — realistic numbers for a hard, imbalanced problem.
- AUC
- 0.91
- F1 score
- 0.69
Highlights
- Risk-ranked customer lists
- Full-metric honest reporting
- Thresholds tuned to the workflow
