About

BuilderFlow is a comprehensive machine learning analysis platform on the Zerve OS that identifies user activation and retention drivers for the Zerve platform itself. The canvas ingests 226.6K behavioral events from 1,472 users during their first 7 days and builds a calibrated XGBoost model (PR-AUC 0.269, Lift@10% 2.56×) to predict 30/90-day retention and 60-day upgrades. The pipeline engineers 51 leakage-free features across 6 dimensions (intensity, breadth, adoption, session stats, collaboration, metadata), segments users into 6 behavioral archetypes via KMeans clustering (silhouette 0.52), and uses SHAP TreeExplainer to identify that "active_days" (SHAP 0.229) is the strongest retention driver. A full exploratory analysis reveals retention patterns by behavior; propensity score stratification estimates causal impact of early behaviors (execution, agent use, collaboration); and an uplift estimation framework ranks 4 product interventions (Agent→Block UI conversion, Day 1/3/7 email drip, session milestone checklist, onboarding→build nudge) by composite priority score accounting for addressable audience, expected uplift, engineering cost, and speed-to-market. The deployable Scheduled Job layer re-scores all users daily, assigning risk tiers and exporting calibrated retention probabilities for the product team to trigger lifecycle campaigns. Key finding: the transition from passive exploration (AI-only sessions) to active construction (block execution, canvas creation) within the first 7 days separates 81% retention "Hands-On Builders" from 4% retention "Casual Browsers": bridging this gap is the core lever for user activation and monetization on Zerve.