Your code review culture is your engineering culture.
Pull request behavior is the highest-resolution signal of psychological safety, seniority distribution, and delivery health on an engineering team. Most leaders look at it last.
I've audited engineering orgs from 12 to 1,400 people. The single best leading indicator of team health isn't sprint velocity, isn't engagement scores, isn't even attrition. It's the median time-to-first-review on a pull request, paired with the tone of the comments. Both are public, both are measurable, and both are nearly impossible to fake.
- Reviews start within hours, not days.
- Comments ask questions before asserting answers.
- Senior engineers explain the 'why,' not just flag the 'what.'
- Approvals are distributed across the team, not bottlenecked on two people.
- PRs sit 3+ days waiting for a single reviewer.
- Comments are nitpicks and tone debates.
- Senior engineers rewrite the code in review instead of teaching.
- Junior engineers stop opening PRs because the review experience hurts.
- Median and P90 time-to-first-review, by team.
- Review participation: % of engineers who reviewed at least one PR per week.
- Comment-to-approval ratio. Above 8:1 is usually a tone problem, not a quality problem.
- % of PRs reviewed by someone two or more levels junior — your real mentorship metric.
Amy Edmondson's psychological safety research shows that team performance correlates more strongly with whether members feel safe to take interpersonal risks than with almost any other variable. The pull request is the highest-frequency, most-public interpersonal risk on an engineering team. Every hostile comment, every multi-day silence, every senior rewrite-instead-of-teach is a small public signal of what's tolerated. Junior engineers read those signals correctly within 4 weeks.
Add James Q. Wilson's 'broken windows' theory: small visible signs of disorder normalize larger ones. A team where 3-day review delays are tolerated will, within 6 months, tolerate 7-day delays. A team where one nitpick comment goes unchecked will tolerate ten. PR culture is the broken windows layer of engineering culture — easy to ignore, expensive to fix once it's bad.
A 60-engineer platform team, as one HR leader recounted, in 2024 had a 4.5-day median review time and rising attrition. They didn't change tooling — they instituted three rules: a 4-hour SLA on first review, a 'questions before assertions' comment style, and a weekly review-tone retro. Within 90 days, time-to-first-review was 3.2 hours, comment-to-approval ratio dropped from 12:1 to 4:1, and three junior engineers — who'd been quietly disengaged — opened PRs 3× more often. Attrition stopped.
- Median and P90 time-to-first-review, per team.
- Comment-to-approval ratio. Anything over 8:1 is usually a tone problem.
- % of engineers who reviewed at least one PR per week.
- % of PRs reviewed by someone 2+ levels junior — your real mentorship signal.
- Publish these numbers in the weekly engineering review. Sunlight changes behavior.
- Make 'first reviewer assigned within 1 hour' a team norm, not an aspiration.