Reading order for experiment results

Reading order for experiment resultsReading order for experiment results: split check, primary metric, guardrails, planned segments, decision, with a stop if the split check fails1. SRM checkis the split sound?2. Primaryeffect and interval3. Guardrailscost, latency, errors4. Segmentsplanned ones only5. Decidesurprising? recheckFails? Stop. Find the cause.Report no effects.

Read every experiment result in the same order, and brief the AI harness to use it too. A failed split check ends the analysis, however good the primary metric looks.