Where our code review time went
After becoming a tech lead, I studied our team’s code review history: hundreds of pull requests and more than a thousand review comments. I wanted to understand where delivery was slowing down.
Much of the elapsed time was spent waiting for a reviewer. Few pull requests received a formal request for changes, but almost every one went through the same queue.
What the data can tell us
Low rates of requested changes don’t prove reviews are unnecessary. A review may confirm a sound approach, catch something in a comment, or help a teammate understand the code. But the waiting time gave us a reason to look at which changes needed that process.
Two proposals
- Match review depth to risk. A disposable experiment and a first production launch need different checks. Define the expectations for each kind of work so that the team knows when to ask for a review and what the reviewer should examine.
- Make coding standards easier to apply. Keep a short core of rules, scope the rest to the files they concern, and automate the checks that CI can enforce. That leaves more review time for correctness, design, and operational risk.
These are proposals the team is still considering. We haven’t measured their effect on delivery time.
What the analysis doesn’t show
The review describes the current queue. It doesn’t establish how a different review policy would affect defects or rework. That would need to be measured after adoption.