Experiment Decision Record
Use this for a bounded change where the team expects evidence to determine what happens next.
Before rollout
| Field | Record |
|---|---|
| Decision this test informs | |
| Current observation and source | |
| Hypothesis | If we…, then…, because… |
| Audience, service, market, or cohort | |
| Change being tested | |
| Baseline and comparison period | |
| Primary measure and definition | |
| Guardrails | Trust, cost, quality, capacity, consent, safety |
| Expected signal and maturation time | |
| Minimum evidence required | |
| Decision rule | Scale / Iterate / Hold / Stop |
| Owner and supporting functions | |
| Rollback or fallback |
After rollout
Record the implementation date and evidence that the intended change actually went live. Then capture the result, confidence, caveats, qualified-outcome evidence, unexpected effects, decision, owner, and next review.
Keep these states distinct:
idea → recommendation → adopted decision → implemented change → observed evidence → verified result
A saved campaign, enabled workflow, published page, or approved draft proves implementation only when the live behavior is checked. It does not prove the expected business result.
Stop or repair the test when the comparison is invalid, tracking changed, trust guardrails fail, the audience was not the intended cohort, capacity makes the outcome undesirable, or the evidence will not mature in time to support the decision.
Use with Measurement Trust and Decision Brief.