A review you run for twenty minutes every week beats a review you run for two hours once a month. The pattern-finding gets easier with more frequent, smaller passes, and a short routine is one you'll actually keep doing once the initial motivation wears off.
The routine
Minutes 0–5: log anything you haven't yet
Before reviewing, make sure the week's practice sessions and any practice-test mistakes are actually in the log. Reviewing an incomplete log just teaches you the wrong pattern.
Minutes 5–10: count errors by type for the week
Filter to just this week's entries and count how many fall into each error type. You're not looking for anything subtle yet — just which type has the most rows.
Minutes 10–15: check it against last week
Compare this week's most common error type to last week's. If the same type shows up two weeks running, that's the one worth acting on — a single week can be noise, but two in a row usually isn't.
Minutes 15–20: write one line for next week
End with a single, specific sentence — not a vague intention. "Redo the last five geometry mistakes before Thursday" is useful. "Focus more on geometry" is not; it doesn't tell you what to actually open on Thursday.
Why weekly beats "whenever I remember"
A pattern that takes four practice tests to show up under a monthly review shows up after two weeks under a weekly one, because you're checking more often with less data required each time. The habit is also easier to keep — twenty minutes on a set day is a much smaller ask than a two-hour catch-up session that keeps getting pushed to next weekend.
Pick a day and treat it as fixed
Sunday evening, or whatever day you review test results, works better than "sometime this week" — a routine without a fixed slot quietly stops happening within a month. If you're prepping for a specific test date, work backward and pick a day that leaves you at least one weekday before your next practice test to act on what you find.
For what to actually do once a pattern is clear across a few weeks of these reviews, see turning an error log into a study plan.