About

This project analyzes user behavior data to identify which actions and workflows are most predictive of long-term user success.
Using event-level logs (notebook runs, deployments, activity timestamps, and upgrade indicators), the data is transformed into user-level behavioral features such as activity duration, engagement intensity, and deployment frequency.
A clear definition of success is used: a user is considered successful if they upgrade to a paid plan or remain active for at least 30 days.
Zerve’s built-in AI Agent is used to automate feature engineering, train an interpretable machine learning model, and rank behaviors by their predictive importance.
The results highlight that early deployments, sustained activity over time, and higher engagement are the strongest indicators of long-term success. These insights can help improve onboarding, retention, and product adoption strategies