Defend your readout against follow-up grilling: SRM, peeking, and the decision.
1. SRM first
Sample-ratio mismatch invalidates everything downstream. State your SRM check (chi-square on split counts) and its result before any metric talk. No SRM check = no readout.
2. Peeking discipline
If anyone looked mid-flight without a sequential/spending plan, say so and quantify the inflation risk. Defending a peeked p-value as clean fails the mock.
3. Readout structure
Decision first (ship / iterate / kill), then primary metric + CI, then guardrails, then segments, then threats (SRM, peeking, novelty, seasonality). End with the one experiment you would run next. WE W2 is the reference bar.