Frame a business problem as a machine learning project

Help me frame a business problem as a machine learning project before I build anything. Business goal: [e.g. "reduce churn among annual subscribers"] How a prediction would be used: [who acts on it, when, and what they do differently] Data available: [sources, history length, rough volume, label availability] Constraints: [latency, interpretability needs, regulation, team skills, budget] Work through: 1. Is ML needed? Could rules, a heuristic, or a simple analysis solve this? Say so if yes. 2. Prediction target: the exact label definition, the unit (customer, order, session), the prediction time, and the outcome window. Flag label leakage risks and ambiguous cases. 3. Problem type: classification, regression, ranking, forecasting, clustering - and why. 4. Metrics: an offline model metric that fits the use (e.g. precision at the top 10% if only 10% can be contacted) and the business metric it should move. Explain the cost of false positives vs. false negatives. 5. Baseline: the simple benchmark the model must beat. 6. Data requirements: features likely available at prediction time, class balance, and how to split data (time-based if appropriate). 7. Risks: feedback loops, fairness or bias concerns, drift, and what happens when the model is wrong. 8. Proof of value: how to test impact (pilot, holdout group, A/B test). Return a one-page project framing doc with these sections, plus a go / no-go checklist and open questions marked [CONFIRM].