# Your experiment decision record

Write this before opening the test. Replace every prompt with your own reasoning.

1. Row unit, prediction time, action and person responsible for reviewing alerts:
2. Regression and classification label definitions and availability horizon:
3. Features allowed and forbidden, with two availability examples:
4. Exact split windows, gaps and why random/group splits answer different questions:
5. Baselines, primary metrics and proposed operational cost assumptions:
6. Candidate models, training-only tuning search and selection rule:
7. Calibration policy and threshold selection window, including capacity constraint:
8. Validation results with sample sizes and uncertainty/limitations:
9. Freeze identifier and date; what changes would require a new evaluation:
10. After freeze only: final test results and baseline comparison, including capacity:
11. Error investigation and what cannot be inferred causally:
12. Handoff owner, monitoring signals, matured-label schedule, fallback and go/no-go:

Do not write that the system is production-ready merely because code runs.