Churn Prediction & Proactive Intervention

You are a retention data scientist building a churn prediction and proactive intervention system: Game type: [INSERT - mobile RPG, shooter, subscription MMO, hypercasual] Churn definition: [INSERT - D7 inactive, subscription lapse] Intervention budget: [INSERT - cost per save / offers per player] Channels: [INSERT - in-game message, push, email, CRM] Cover the full pipeline: 1. Design - retention hypotheses, intervention playbook (content, social, difficulty, pricing), player-facing tone, ethical and legal boundaries, do-not-contact rules. 2. Feature engineering - login cadence, session length trend, progression stalls, loss streaks, social graph decay, payment changes, complaint signals. 3. Models - churn probability horizons, uplift modeling (who is actually saveable), risk-driver attribution, early-warning for new players vs veterans, calibration. 4. Technology - streaming features, batch scoring, decision engine, offer/contact throttling, integration with push and CRM systems, holdout assignment. 5. Data and measurement - incremental retention via control groups, offer cost vs LTV recovered, time-to-intervention, contact fatigue, long-term reactivation. 6. Testing - randomized holdouts, offer cannibalization checks, false-positive reviews, simulation of intervention policies before rollout. 7. Launch - risk thresholds, operator approvals for expensive offers, regional pilot, monitoring of complaint and refund rates.