It's a checkbox step, not a real review.
Reviews that go through the motions
A review step exists, but it's too fast, too routine, or too deferential to catch real problems. The presence of a review creates confidence that the work was checked, even when little checking actually happened.
- Nearly everything that reaches the review gets approved.
- Reviews take a few minutes for work that would take an hour to check properly.
- Mistakes the review should have caught get through.
Who: the person who runs the review step. Look at: the last quarter of reviews. Count: the approval rate, and how many times the review stopped or changed something. If you can, plant one known mistake in a piece of work before it goes to review, with the reviewer's agreement in advance, and see whether it gets caught. You're probably looking at this pattern if nearly everything passes and the planted mistake gets through. Probably not if the review regularly sends work back and catches the planted mistake.
Checking can't keep up. In that pattern, reviewers want to check and don't have time. In this pattern, the review has turned into a ritual whether or not there's time.
On an important piece, name one person who is not the author to verify each key claim against its source.
Name one person who signs off before a category of work goes out.
On an important hire or decision, give one outsider a real say in whether it meets the bar.
An approval rate near 100% at a sign-off step usually means the step has stopped checking. Observation. People asked only to watch over automated work get worse at spotting its errors over time (Bainbridge, 1983). Our reasoning, built on a well-known argument from human factors research.