About

🚀 Built an AI-Powered Customer Retention System for the Zerve AI Hackathon! 🚀

ChurnGuard Pro predicts which customers will churn 30-60 days BEFORE they leave - giving businesses time to save them.

📊 THE RESULTS:
✅ 87.21% accuracy (ROC-AUC)
✅ 57% recall - catches most churners early
✅ $34.8M annual net benefit for 10K customers
✅ Validated with 5-fold cross-validation

🎯 THE INNOVATION:
Most solutions just say "will they churn?"

ChurnGuard goes further:
- WHY they'll churn (risk factor analysis)
- PRIORITY scoring (risk × customer value)
- PERSONALIZED recommendations (what to do about it)
- Risk levels (🔴 High / 🟡 Medium / 🟢 Low)

💡 REAL RESULTS:
Test Case 1: 98% churn risk flagged → $450K customer → Immediate intervention
Test Case 2: 11% risk → Loyal customer → Efficient resource allocation

🔧 TECHNICAL HIGHLIGHTS:
- 25 engineered features (age_risk_score, engagement_score, CLV)
- Ensemble model: Logistic Regression + Random Forest + Gradient Boosting
- Production-ready API with error handling
- Built entirely in Zerve AI in 5 days

📚 KEY INSIGHTS:
1. Customers with 3-4 products = 100% churn (complexity kills!)
2. Inactive members = 2x churn rate
3. Feature engineering > fancy algorithms

This is my FIRST end-to-end ML project as a Business Analysis student. Combined business strategy + data science + production thinking.