A pull request’s open-to-merge time is elapsed time, not the amount of work reviewers spend reading code. It can include waiting for a first human response, review and revision rounds, CI runs, and time between approval and merge. To find out why your team’s PRs are slow, measure those stages separately before changing the workflow.
What counts as pull request review time?
“Review time” can mean either active effort or total elapsed time. Elapsed time runs from submission to merge; it may include long periods when nobody is reviewing or editing the change. A useful stage map is:
As an Amazon Associate I earn from qualifying purchases.
- Submission to first human response: time until a reviewer meaningfully engages. A bot acknowledgment is not the same as a human review.
- Review and comments: time spent examining the change and communicating findings.
- Author revision: time to assess comments, make changes, and update the PR.
- Further review: time for reviewers to check the updated diff and resolve outstanding questions.
- Approval to merge: time after acceptance, including any remaining checks or coordination.
CI and build duration can overlap with these stages, so record it separately rather than assuming all elapsed time is human waiting. A single open-to-merge number cannot tell you which stage is responsible.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsHow to measure where the time goes
Use timestamps from your repository or review system to calculate elapsed time for each stage. Track both the median and a tail measure such as the 75th percentile (P75): the median describes a typical change, while P75 helps expose delays that affect a substantial portion of PRs without letting a handful of extreme cases dominate the picture. Meta says it tracks P75 Time In Review; that is one company’s operational choice, not a required standard.
#1 Best Overall
- Separate bot first responses from human first responses.
- Count comment-and-revision cycles, and measure the time between each reviewer comment and author update.
- Record CI wait and run duration, along with build results.
- Measure approval-to-merge time independently.
- Compare results by change size, project, or other relevant context, rather than treating every PR as equivalent.
Use a consistent definition for each timestamp and compare the same kind of work over time. A stage-level breakdown turns “PRs are slow” into a testable diagnosis.
What evidence says about common delays
The evidence supports investigating waiting and repeated review cycles, but it does not establish that most PR time everywhere is waiting and rework. Results vary by stage, platform, and project.
Delays change as a PR moves through review
A 2021 study modeled 3,347,937 pull requests from 11,230 GitHub projects and found that important latency factors differed across stages and contexts. At submission, factors included description length, code changed, developer experience, reviewer activity, and project workload. For PRs with comments, first-comment latency and comment count mattered; for PRs using CI, build latency and build results mattered. These are observational findings, not proof that any one factor causes delay in every team. Read the study.
First response needs to mean human engagement
A 2023 study of 111,094 closed PRs from ten popular GitHub open-source projects analyzed bot-generated and human-generated first responses separately. That distinction matters operationally: an automated acknowledgment may confirm that a PR was received, but it does not show that a reviewer has taken responsibility for it. Read the study.
Approval does not always mean the work is finished
A 2022 analysis of about half a million reviews hosted on Gerrit and Phabricator distinguished waiting before the first response from waiting after acceptance and before merge. In the sampled Phabricator projects, post-acceptance waiting accounted for 29%–63% of total review lifetime. The reported median post-acceptance share was 43.00% in Phabricator and 1.00% in Gerrit. The samples differ, so this is not evidence that one platform causes faster merging, nor is it a universal share for all teams. It does show why approval-to-merge time is worth measuring. Read the study.
Match the fix to the stage that is slow
Treat a suspected cause as a hypothesis, then check whether the relevant stage improves after a change. The same intervention can have different effects across teams, and faster review is not useful if it undermines review quality.
Rank #4
| What the data shows | What to investigate | Possible team response |
|---|---|---|
| Long submission-to-human-response time | Reviewer ownership, workload, or routing | Make ownership explicit; check whether the requested reviewers are available and well matched to the change. |
| Many comment-and-revision cycles | Change clarity, diff size, or feedback arriving late | Improve the PR description, split a very large change where practical, or seek design feedback earlier. |
| Long CI waits or runs | Build duration, reliability, or failing checks | Measure queue time and execution time separately; address the build bottleneck indicated by those measurements. |
| Long approval-to-merge time | Remaining checks, merge ownership, or manual steps | Clarify who owns the merge and whether the post-approval path can be made lower-friction. |
Improve response speed without sacrificing focus
Prompt acknowledgment helps authors know their change has an owner, but asking reviewers to interrupt focused work constantly can trade one delay for another. Google Engineering Practices frames review speed as a way to improve the whole team’s flow, while also emphasizing code health. Its guidance says, “One business day is the maximum time it should take to respond to a code review request (i.e., first thing the next morning).” That is Google’s internal guidance, not a universal service-level standard. See Google’s review-speed guidance.
If a reviewer cannot complete a review promptly, Google suggests communicating when they expect to review, suggesting another reviewer, or providing broad initial feedback at a suitable break point. That gives the author useful information without implying that every review must interrupt the reviewer immediately.
Best Value
Make large changes reviewable
Google’s guidance recommends asking for a very large change to be split into smaller changes that build on each other when practical. Smaller, coherent diffs can make review easier to schedule and understand, but splitting is not always appropriate: dependencies or the risk of leaving an incomplete change may make a single review more sensible. Choose a boundary that preserves correctness and context, not an arbitrary line-count target.
Make the next action clear
A review should leave the author and reviewer able to tell what happens next. State whether a comment blocks approval, distinguish a required correction from a suggestion, and make ownership of follow-up or merging visible. Google also advises reviewers to communicate clearly about minor unresolved comments when approving; that can prevent an approval from leaving uncertainty about whether work remains.
What company examples do—and do not—show
Meta reported that a change to reviewer recommendations increased the share of diffs reviewed within 24 hours by 1.5 percentage points, while the accuracy of its top-three reviewer suggestions rose from below 60% to nearly 75%. Meta also describes tracking P75 Time In Review and using a stale-diff nudging bot. These are reported outcomes and practices at Meta, not guaranteed results for another organization. Read Meta’s account.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Google Research reported that, in its deployment, authors addressed 7.5% of reviewer comments by applying an ML-suggested edit. That is an adoption figure, not proof that AI tools reduce review duration overall, improve every suggestion, or replace reviewer judgment. Read the Google Research paper.
Quick Recap
Run a small workflow experiment
- Choose a stage metric, such as median and P75 time to first human response, revision-cycle time, CI duration, or approval-to-merge time.
- Choose one intervention aimed at that stage, such as clearer reviewer ownership or a more explicit merge handoff.
- Keep the metric definition consistent and compare a suitable period before and after the change, accounting for differences in PR types or workload where possible.
- Check quality and team experience alongside speed: for example, whether follow-up review, missed issues, or disruptive interruptions increased.
- Keep, adjust, or revert the intervention based on the local results rather than assuming a practice that worked elsewhere will work the same way.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
