Coverage begins with the contract
Unit, integration, end-to-end, accessibility, visual, and performance tests answer different questions. The PRD should identify the acceptance criteria, while the QA plan chooses the smallest credible set of evidence for each one.
This avoids the common situation where many tests pass but the most important user outcome was never exercised.
Store evidence with revision context
A screenshot, log, trace, or test result should identify the build, environment, requirement revision, and acceptance criterion it supports. When any of those change, the system can explain which evidence is stale.
- Use stable acceptance-criterion IDs.
- Record the tested build and environment.
- Keep failures and waived checks visible.
- Expire evidence when its upstream contract changes.
Release decisions become explainable
Traceable evidence lets a release approver answer a practical question: what do we currently know about this build? The answer is not a vague confidence score but a set of passed, failed, missing, and consciously accepted checks.
That makes QA a source of product knowledge instead of a gate that appears only at the end.