阅读中文版本

Review Metrics and Quality Workflow

Use review metrics such as turnaround and rework to find process bottlenecks without gaming reviewers.

Overview

Improve the review process using metrics that reveal bottlenecks, not vanity numbers.

This workflow is reusable across teams and environments; adapt the commands to your tooling while keeping the sequence and checkpoints intact.

Environment and prerequisites

• Review analytics
• Pull request data
• A defined set of metrics

Procedure

1. Choose meaningful metrics

Select metrics like review turnaround, change size, and rework rate.

Expected state: Metrics reflect process health, not individual scores.

2. Find bottlenecks

Identify where changes wait, such as review assignment or approval.

Expected state: The biggest delay is identified.

3. Fix the process

Address the bottleneck with process changes such as reviewer rotation or smaller changes.

Expected state: The targeted delay shrinks.

4. Watch for gaming

Check that improvements are real and not the result of gaming the metric.

Expected state: Metrics drive genuine improvement.

# review turnaround distribution
gh pr list --json createdAt,mergedAt | jq -r '.[] | .createdAt + " " + .mergedAt'

Verification

Confirm the targeted bottleneck improved and metrics are not being gamed.

Decision points

Measure individuals or the process?

• Process metrics only
• Per-person metrics

Recommendation: Measure the process; per-person metrics incentivise gaming and fear.

Optimise turnaround or thoroughness?

• Balance both
• Minimise turnaround

Recommendation: Speed up reviews without reducing rigour; both matter.

Common failures

• Per-person metrics cause gaming and anxiety.
• Optimising speed alone reduces review quality.
• Unmeasured bottlenecks persist.