About

This project analyzed 409,287 events from 5,434 Zerve users over 98 days to identify behavioral predictors of long-term success. Success was operationalized as users returning 7+ days after signup AND using core features (code runs or agent calls) within their first 14 days — 206 users (4.3%) met this definition.
Key Findings:

Feature breadth matters most: Exploring 3+ distinct platform areas by day 3 is 1.7x more predictive than agent adoption alone
Agent-first adoption: 100% of successful users interact with the agent; 61% start there
Day 3 is critical: Users inactive by day 3 rarely return

Prediction Model: Logistic Regression (AUROC 0.8278) identified 3 actionable decision rules to identify high-potential users before day 7, enabling early intervention. Conclusion: Success comes from breadth of exploration, not depth in single features.