About

Analyzed behavioral data from 4,774 Zerve users across 3 months
to find which early user behaviors predict long-term retention.

Key finding: 97.8% of users churn within 30 days. But retained
users show 33x more events, 68x more sign-ins, and 278x more
code runs than churned users.

Built a Random Forest model (98% accuracy) on 15 behavioral
features. Top 3 predictors: Days Active, Sign-in Count,
Week 1 Events.

Recommendation: Zerve has a first-week problem. Users who run
even one code block in week 1 are dramatically more likely to
stay long-term.