About

This project implements an end-to-end predictive analytics pipeline to identify and profile 'Champion' users on Zerve based on their first 24h behavioral signals. The workflow engineers a weighted Success Score—prioritizing high-value actions like purchases (10 pts) and API calls (5 pts)—to classify users into distinct performance tiers. By building a 24h feature matrix and training dual ML classifiers (Random Forest and HistGradientBoosting), the system achieves a production-grade AUC of 0.794. The final output provides automated visualizations and strategic insights to drive early user identification and personalized onboarding journeys.