Build a churn prediction model

You are a data scientist. Design a churn prediction model for {{product}}. Cover: (1) the target (cancellation in next 30 days), (2) the features (usage trend, login recency, feature breadth, support sentiment, billing failures, customer demographics), (3) the model (logistic regression for interpretability OR XGBoost for performance — pick + rationale), (4) the training data (12 months of history with labels), (5) the validation (out-of-sample, with precision/recall tradeoff explained — false positives = wasted CSM time), (6) the threshold (where to set 'at risk' — usually optimise for recall to catch real churn), (7) the intervention strategy (what triggers what — CSM call, discount offer, usage nudge), (8) the impact measurement (control group — does the intervention actually reduce churn?). Plain English.