Auto-merge rate measures the percentage of pull requests integrated automatically instead of waiting for human review. Treat it as a throughput metric, not proof that automation is safe: pair it with escape rate, test reliability, and merge-queue wait time. A rising rate is valuable when it removes repetitive approvals without increasing production defects. section 10.7.1 section 10.5.2
Raise the rate only when the policy and evidence justify it:
- Require passing tests, sufficient coverage, architectural lint compliance, no new code smells, a bounded diff, and a bounded blast radius. section 10.5.2
- Exclude Red-flagged surfaces such as authentication, payments, and data schemas from automatic merging. section 10.5.2
- Promote patterns gradually, beginning with conservatively selected changes and waiting for 50 consecutive clean approvals before making a pattern Green. section 10.5.2
- Track the rate as a target rather than a ceiling; a flat Green bucket over two quarters signals stalled automation rather than success. section 10.5.2
A measurable maturity path
A team began with an auto-merge rate of 22%, reached 34%, 48%, and then 61% over six months by promoting proven approval patterns. At 61%, the remaining Yellow and Red reviews fit comfortably into the reviewers’ capacity, while escape rate remained 0.3%, below a 3% tolerance. The improvement reduced review-queue pressure without treating automatic merging as permission to weaken safeguards. section 10.7.1 section 10.5.2
Listing 10.4 Policy-based ownership with triage rules
libs/core: owner: platform-team green: "coverage > 95%, diff < 50 lines, no new dependencies" yellow: "platform-team-oncall reviews" red: "platform-team-lead + security-review" services/checkout: owner: payments-team green: "tests pass, no schema changes, no auth modifications" yellow: "payments-team-member reviews" red: "payments-team-lead reviews" sla: "99.99% - minimum Yellow review for all changes" services/notification: owner: platform-team green: "tests pass, no external API changes, diff < 100 lines" yellow: "any platform-team member reviews" red: "platform-team-lead reviews"
Use policy-based ownership rules to make Green, Yellow, and Red decisions explicit for each code area, including coverage, diff size, schema changes, and review ownership. section 10.5.2
