New engineering managers often arrive asking for dashboards. What they need first is three numbers, four conversations, and the patience to wait two weeks before changing how the team works.
Three numbers, in order
Start with review latency — the cheapest signal of collaboration vs queueing. Then PR size distribution — large PRs usually mean missing structure, not missing effort. Only then cycle time.
If reviews are slow and PRs are huge, cycle time will only yell louder. Fix the stage that moved first.
Four conversations, then act
Pair every metric with a conversation. Ask the top reviewer if load felt fair. Ask the author of the largest PR if it could ship in pieces. Ask the tech lead which rituals they protect. Ask the quietest engineer what is in the way.
Wait two weeks before changing process
Most things that look broken in week one are explained by context you have not earned. Resist rewriting standup, retro, or the PR template until you understand why they exist.
When you do change something, change one thing. Name the metric you expect to move and revisit in a month. That builds more trust than any dashboard.
Key takeaways
- Review latency, PR size, then cycle time — in that order.
- Four conversations turn metrics into trust, not surveillance.
- One process change at a time, with a named expected outcome.
