| name | release-quality-plan |
| description | Define the quality evidence required for a release by mapping changed behavior, critical journeys, risks, environments, test ownership, entry/exit criteria, known gaps, and post-release validation. |
Release Quality Plan
Use when a release or milestone needs an explicit quality gate rather than relying on accumulated ad hoc tests.
Procedure
- Define release scope, changed surfaces, critical user/system journeys, supported platforms/environments, migrations/config changes, and consequences of failure.
- Review implementation and architecture risk assessments, known defects, incident history, dependency changes, and unresolved assumptions to build the release risk picture.
- Map each material risk to evidence and owner: static/build checks, unit/integration/contract/E2E, exploratory, accessibility, security, performance/load, migration/rollback, operational smoke, or other validation.
- Define environment/data requirements and what each environment can legitimately prove. Call out gaps where production-like behavior cannot be reproduced pre-release.
- Define entry criteria for release validation so QA does not test a moving or incomplete target unknowingly.
- Define exit criteria around evidence and residual risk rather than “all tests pass.” Include blocking defect thresholds, required independent reviews, and accepted exceptions with owners.
- Plan regression coverage proportional to blast radius, shared surfaces, and preserved contracts, including compatibility/mixed-version states when releases roll incrementally.
- Define post-release checks and monitoring needed to catch defects that only appear with real traffic, data, platform, or distributed placement.
- Track evidence by exact release revision/artifact so results cannot be silently reused after meaningful code/config changes.
- Publish a concise quality disposition: ready, ready with accepted residual risk, or not ready, including missing evidence and the owner/decision required.
Decision rules
- Release quality is a risk decision supported by evidence, not a raw test-count or pass-rate calculation.
- Do not treat lower environments as proof of behaviors they do not reproduce.
- Quality ownership stays independent from implementation; accepting residual product/business risk belongs to the appropriate decision owner.
- In Fleet deployments, include distributed/node/placement validation only when it is material to the release, while keeping live scheduling policy Fleet-owned.
Quality gate
The plan is complete when release risks and critical journeys are explicit, each material risk has an appropriate validation owner/evidence path, environments and revision identity are clear, exit criteria expose residual risk, and post-release verification can detect failures that pre-release testing cannot prove away.