Metrics · Updated 2026-09-17
Review latency
The time between a pull request being ready and a reviewer responding to it — usually the largest single component of cycle time.
Also called: Time to first review · Review turnaround
What it is
Review latency measures waiting, not working. A change sits ready while its author context-switches to something else, and the cost compounds: by the time feedback arrives, the author has to reload the whole problem into their head before they can respond to it.
Measure time to first response separately from time to approval. They have different causes and different fixes — the first is about attention and routing, the second is usually about pull request size.
Why it is the metric AI review most plausibly improves
An automated reviewer responds in minutes, at three in the morning, on a Friday. That does not replace human review, but it does change what the author is waiting for: obvious problems surface while the change is still fresh, and the human reviewer arrives to a cleaner change.
This is also the most honest ROI story for the category. Claims about bugs caught are hard to verify; a drop in time-to-first-feedback is measurable in your own platform data, before and after.
Why it matters when you are evaluating
Baseline it before the trial and measure the same window after. If a vendor offers analytics, check whether they report time to first human response as well as bot response — a tool that comments instantly while humans take just as long has moved a number without moving the outcome.
Common mistakes
- Celebrating faster bot response while human review latency is unchanged.
- Measuring the mean. Latency distributions have long tails, and the tail is what people remember; use the median and the 90th percentile.
- Ignoring the weekend effect, which can dominate a weekly average.