Key takeaways
- Start the clock when a change is ready for review, not when the author first pushed it.
- Report first response and full review separately so neither number hides the other.
- Set a standard lane and a high-risk lane instead of one target for every change.
- Review the weekly distribution with the pod and fix chronic causes rather than naming individuals.
- Treat the target as a queue-management signal, never as evidence of review quality.
Why a target beats a vague expectation
A review queue without a target is governed by whoever is busiest at that moment. People learn to expect an answer sometime today, or maybe tomorrow, and they plan around that uncertainty. A written target turns the expectation into something the team can inspect and improve. It also gives the reviewer permission to protect time for review instead of treating it as leftover work.
A target does not need to be aggressive to be useful. The first value of the number is that it makes waiting visible, not that it forces speed. Teams often discover that most of the delay sits in handoffs rather than in reading code. Once that is visible, the pod can fix the queue instead of asking reviewers to try harder.
Start the clock at the right moment
Measure from the point when the change is ready for review. That means the description is complete, the checks have run, and the author has named the reviewer or the review group. Measuring from the first push counts unfinished work as waiting time and punishes authors for preparing properly. It also makes the metric impossible to compare across changes.
Record the moment of the first substantive reviewer response and the moment the review is complete. A short acknowledgement is not a review, and it should not stop the clock. If the pod needs an acknowledgement target, keep it separate from the review target. Mixing the two lets a fast hello look like a fast decision.
Separate author time from reviewer time
Split the elapsed time into waiting for a reviewer, active review, and author rework. Each part has a different owner and a different remedy. Long waiting points to capacity or routing, while long active review points to change size or missing context. Rework points back to the brief, the tests, or the definition of done.
Report the first response time and the full review time as two numbers. A pod can answer quickly and still take days to finish, and a single average would hide that pattern. Two numbers also help the buyer see where its own decision delay enters the process. Reviewers cannot fix a queue that waits on product answers.
Pick a target the pod can influence
Set the target from the pod's own recent distribution rather than from an industry benchmark. Look at the median and the slowest quartile for normal changes, then choose a target slightly better than the current median. A target that requires a structural change in staffing will be ignored or gamed. A modest improvement is more likely to hold.
Use two lanes: a standard lane for small, well-understood changes and a high-risk lane for migrations, security work, and shared infrastructure. The high-risk lane gets a longer target and a named senior reviewer. This keeps the fast lane honest without encouraging shallow review of the changes that matter most. Review the lane definitions each quarter so a change that becomes routine can move to the faster path.
Make room for deep review
A fast target can reward a reviewer who skims and approves. Prevent that by naming what a complete review includes: the behavior change, the boundary, the evidence, and the rollback. A review that does not answer those questions is not complete, no matter how quickly it was submitted. The target applies to complete reviews only.
Give reviewers a way to stop the clock when the change is not ready. If the description is missing or the checks have not run, the reviewer returns it with a specific request. That return is not a review and should not count against the target. It also teaches authors what a ready change looks like.
Publish the trend without ranking people
Show the weekly distribution to the whole pod, including the median, the slowest cases, and the reasons recorded for delays. Do not publish a leaderboard of reviewers. Ranking individuals on turnaround encourages fast approvals and discourages people from taking the hard changes. The point is to improve the queue, not to grade people.
When a case is slow, ask what the reviewer was waiting for. Common answers include missing context, an unresolved product question, or a change that was too large to read in one sitting. Those answers lead to fixes in the brief, the decision path, or the size limit. A target that produces those fixes is doing its job.
Limitations
Turnaround time says nothing about whether the review was good. A fast review of a risky change can be worse than a slow one. Pair the target with defect escape data and with spot checks of review comments. Neither metric is sufficient alone, and the team should resist treating either as a score.
Small pods have noisy distributions, so a single slow week can look alarming. Use a rolling four-week view before drawing conclusions. Also remember that the buyer's own decision speed sits inside the elapsed time. A target that ignores the buyer's side will blame the wrong team.
Handle exceptions without breaking the rule
Some changes cannot meet the standard target, and the process needs a way to say so without inviting abuse. Name the conditions that justify an exception: a security fix, a change that depends on an external review, or a migration that needs a maintenance window. The author records the condition in the review ticket and the reviewer confirms it. An exception that is written down is easier to review than a quiet delay.
Keep the exception list short and review it monthly. If one condition appears every week, it is no longer an exception and the target should change. If exceptions cluster around one reviewer, the problem may be capacity or routing. The exception record is a signal about the rule, not a loophole to be closed. Treat it as feedback and adjust the design.
Review the target each quarter
Look at the quarter's distribution before changing the number. If the median improved and the slowest quartile did not, the queue may be healthier for small changes while risky ones still wait. If both improved but defect escapes rose, the target may be pushing reviewers too fast. The review should consider the target, the review depth, and the buyer's decision speed together.
Ask the pod what would make the biggest difference to the next quarter. Common answers are a named reviewer per area, a smaller change size, or a scheduled review window that respects time zones. Pick one change and measure it. A target that stays fixed while the process improves will eventually become either trivial or impossible.
Questions about assessing Philippine developers
What is a reasonable first target?
Start from your own median and aim for a modest improvement. The right number depends on change size, reviewer capacity, and how many decisions wait on the buyer.
Should the target include weekends?
Use working hours for the pod's agreed schedule and state the rule in the metric definition. Elapsed calendar time is easier to collect but harder to interpret across time zones.
How do we stop reviewers from rubber-stamping?
Define a complete review, require evidence, and audit a sample of fast approvals. A turnaround target alone will always reward speed over depth.
