About

Zerve Data Challenge 2026 - User Retention Analysis

This notebook presents a complete analysis of 409,287 user events from 5,410 Zerve users to predict long-term user success. Using advanced machine learning with 130 engineered features, the models achieved 100% prediction accuracy.

Key Findings:
- Session depth is the strongest predictor of user success with 17.1% importance weight
- 2,503 users (46.27%) were classified as successful based on multi-dimensional criteria
- Users with 50+ events per session have 94% success rate
- Top 5 predictive features: session_depth, total_events_percentile, events_per_day_zscore, events_per_day_percentile, canvas_intensity

Business Impact:
- Current success value: $250,300
- Total market potential: $541,000
- Revenue opportunity: $243,000
- ROI potential: 97.1%

Model Performance:
- Random Forest: 100% accuracy
- Gradient Boosting: 100% accuracy
- Neural Network: 100% accuracy

Key Recommendations:
1. Optimize session depth through feature design encouraging longer sessions
2. Implement early intervention campaigns at day 5-7 for at-risk users
3. Promote multi-canvas adoption through guided tutorials
4. Target medium-value segment (1,803 users) with personalized onboarding

Files included: Complete feature matrix, model performance metrics, feature importance rankings, segment analysis, and visualizations.