Write a one-page brief before choosing a model:
| Field | Record |
|---|---|
| Decision | Customer/item, decision time, eligible actions, and default/no-action option |
| Owner and cadence | Who uses the output, batch vs session/API, and when an action can take effect |
| Outcome | Exact event definition, attribution/measurement horizon and reporting delay |
| Economics | Contribution after fulfilment, subsidy, contact, returns and service costs; currency and horizon |
| Constraints | Inventory, minimum margin, subsidy/contact budget, fatigue and applicable customer policy |
| Evidence | Randomised assignments or identification assumptions; historical support for actions |
| Success | Minimum useful effect and uncertainty; service, spend and customer guardrails |
| Fallback | Existing approved policy when features, evidence or serving are inadequate |
Measure contribution instead of revenue, and retained value instead of churn risk. If cost or treatment data is missing, state the narrower predictive question you can answer and collect the missing data.