Keep One Page per Test That Informs a Decision So the Next Person Can Use It
Acceptance Evidence Record template
Companion to The Product Manager’s New Job Is Acceptance · Updated 2026-09-30
The chapter “Keep an Acceptance Evidence Record for Every Test That Informs a Decision” explains the fields. Skip the record if the test would not change a decision.
| Field | Your entry |
|---|---|
| Record ID and date | |
| Versions: product, message, pricing when relevant, ICP | |
| Actor and path stage | |
| Hypothesis (in full), with refutation rule | |
| Expected behavior | |
| Observed behavior | |
| Qualitative evidence: quotes, notes, recordings | |
| Quantitative evidence, with denominators | |
| Interventions: every rescue, by category and actor | |
| Failure and its type: where the path broke, and whether the cause was a defect in the software or a failed assumption | |
| Possible explanation | |
| Competing explanation | |
| Confidence, with a reason | |
| Required change: the smallest change the evidence supports | |
| Retest criteria | |
| Source and provenance: who ran it, where the data is, whether AI drafted or summarized any part |
A completed example (illustrative)
| Field | Entry |
|---|---|
| Record ID and date | AER-014, 2026-10-12 |
| Versions | Product 4.2, message v3, pricing v2, ICP v2 |
| Actor and path stage | Practice manager, setup |
| Hypothesis | A practice manager at a twelve-physician group completes setup from the ad, with the practice’s own scheduling data and no rescue, in one working day. Refuted if five or more of twelve stall. |
| Expected behavior | Setup in under a day; first intake in under five minutes. |
| Observed behavior | Setup took three days. The first intake took eleven minutes on day four. |
| Interventions | Setup rescue by customer success, six hours, unplanned. |
| Failure and its type | The path broke at the schedule export. Failed assumption: the practice knew to export first. |
| Competing explanation | This practice’s schedule data was unusually messy. |
| Confidence | Low: one of a planned twelve practices. |
| Required change | Change the claim to “ready in one week” and build the import step. |
| Retest criteria | The import step ships, or the claim changes again. |
Confidence levels
- Low: a single test, or known confounders that were not ruled out.
- Medium: repeated observations in a small or narrow sample.
- High: consistent results across several segments and conditions.
Write the reason next to the level. “Medium” alone says nothing.
Download