Customer Churn Prediction Analysis

You are a predictive analytics expert. Design a churn prediction project.

Business: [BUSINESS]
Customers: [COUNT]
Churn rate: [%]
Data available: [DESCRIBE]
Goal: [REDUCE CHURN BY X%]

1. Problem Definition:
   - Churn definition: criteria, timeframe, types (voluntary/involuntary)
   - Business impact: revenue at risk, LTV impact
   - Success criteria: model performance, business outcome

2. Data Preparation:
   - Feature categories: usage, engagement, support, billing, demographic, behavioral
   - Feature engineering ideas: 20+ features with descriptions and rationale
   - Data quality checks: missing values, outliers, class imbalance
   - Train/test split: temporal split strategy

3. Model Development:
   - Baseline: simple rules-based approach
   - Models to try: logistic regression, random forest, XGBoost, neural network
   - Evaluation: AUC-ROC, precision-recall, calibration
   - Feature importance: SHAP values, business interpretation

4. Actionable Insights:
   - Risk segments: high/medium/low with characteristics
   - Intervention recommendations per segment
   - Optimal intervention timing
   - Expected impact modeling

5. Production Deployment:
   - Scoring frequency: real-time vs batch
   - Integration: CRM, CS platform, alerting
   - Monitoring: model drift, retraining triggers

6. Measurement: A/B test design, attribution, ROI calculation
7. Presentation: stakeholder deck outline, technical documentation