The GMAT review-and-edit strategy: using your three changes per section

You can change up to three answers in each GMAT Focus section, and most candidates either forget the feature exists or burn their last minutes re-reading questions they got right. The protocol below is the one we teach every client: bank the early questions at pace, flag at most four, keep a fixed reserve, and only change an answer when you can name the error.

The feature is worth real points, but only with a plan. Without one it is a distraction that costs time in the middle of the section and produces panicked edits at the end.

What GMAC actually gives you

The rules come from GMAC (mba.com), and they are tighter than people remember. Within each 45-minute section you can bookmark any question and, before the section clock runs out, return to the review screen, open any question, and change your answer, with a limit of three changes per section (GMAC exam structure). You must answer each question to move forward; there is no skipping. The review screen is not a separate timed period. It runs on whatever is left of the 45 minutes.

Three details decide everything. The clock does not pause. Unanswered questions at the end of a section count against you, so you need an answer on every question before you spend a second on review. And the three-change cap applies to changes, not to looks: you can open a bookmarked question, decide your answer was right, and close it without using a change.

Three changes is a ceiling. It is not a quota.

The per-question pace you have to hold

The review feature only exists for candidates who reach the end with time. That makes pace the foundation of the strategy, and pace is specific. Pace comes first. Quant is 21 questions in 45 minutes, which is 2 minutes 8 seconds each. Verbal is 23 questions, 1 minute 57 seconds each. Data Insights is 20 questions, 2 minutes 15 seconds each.

Those are section averages, and the protocol asks you to run slightly faster than average through the first two-thirds so that a reserve accumulates. The table is the budget we hand clients in week 6.

Bank pace is faster than average by about eight to ten seconds a question. Over 14 questions that is roughly two minutes, and the remaining minutes of reserve come from the questions you flag and move past instead of fighting. The reserve is spent only after the last question is answered.

Bank, flag, reserve: the three-part protocol

The protocol has three parts and they happen in order.

Banking is what you do to most questions. You read it, you work it, you answer with the confidence that you would not want to revisit it, and you move on without a bookmark. A banked question is closed. You will not think about it again, and that discipline is the whole reason the strategy works, because a mind that keeps a dozen questions half-open runs slower on the next one.

Flagging is what you do to a small number of questions where you have an answer but a specific doubt. The key word is specific. “I’m not sure” is not a flag. “I may have read ‘at least’ as ‘at most’” is a flag. “I eliminated to two choices and picked on gut” is a flag. Put an answer down, bookmark it, note the doubt in two or three words on your noteboard with the question number, and move on at pace. Cap it at four flags per section. A fifth flag means you drop the weakest earlier one.

The reserve is the time you reach the review screen with. You spend it on flagged questions in the order of your notes, highest expected gain first: a question where you can check a single misread in 30 seconds comes before a question where you would need to re-solve from scratch. Open the question, check the specific doubt, and apply one of two outcomes. Either you find the error and change the answer, or you do not find it and you leave the answer alone.

Change nothing you cannot explain. That rule has no exceptions.

What a flag-worthy doubt looks like

Most candidates flag badly because they flag feelings. The list below is the set of doubts that justify a bookmark, and if your doubt is not on it, bank the question and move.

• You read a constraint once and did not re-read it before answering (“positive integer,” “distinct,” “at least,” “not”).
• You eliminated to two choices and the choice between them was not reasoned.
• In Quant, your answer was not among the choices until you adjusted something, and you are not sure the adjustment was legitimate.
• In Data Insights, a multi-part question where one part was a guess.
• In Verbal critical reasoning, you identified the argument structure after reading the choices rather than before, which means your prephrase was missing.
• In reading comprehension, you answered a detail question from memory without returning to the passage.
• Any question where you noticed the clock mid-solve and rushed the final step.

That is seven kinds of doubt, and most sections produce two or three of them. Seven is the whole list. If you are producing seven, your banking is too loose and the fix is in pace, not in the review screen.

The time budget inside the reserve

Four minutes sounds like a lot until you open a question. Re-reading a Quant problem takes 30 to 40 seconds before you have done anything. Here is how the reserve divides in practice.

In Quant, with four minutes: open the flag with the cheapest check first (a misread constraint, 30 to 45 seconds), then the second (an arithmetic re-check, 60 to 90 seconds), and leave the third only if two minutes remain. A fourth flag in Quant almost never gets opened, which is why the fourth should be your weakest doubt.

In Verbal, with three minutes: critical-reasoning flags are cheap to check (re-read the conclusion and your choice, 40 seconds) and reading-comprehension flags are expensive (find the lines, 60 to 90 seconds). Check the critical-reasoning flags first regardless of order.

In Data Insights, with four minutes: a single-answer question is cheap; a multi-part set is expensive, often two minutes, because you have to re-read a source. If you have one multi-part flag and two single flags, do the singles first and spend what remains on the set.

Stop with 20 seconds on the clock. Confirm whatever is open.

How to practice the protocol

The protocol fails on test day if it was not rehearsed, because the moment you are most tempted to flag everything is the moment you are tired and behind. We rehearse it in four steps across weeks 6 to 9 of the program, and you can do the same alone.

Step one is pace-only sections. Take a timed 45-minute section with no flagging allowed and the goal of finishing with four minutes left. Most candidates need two or three of these before the bank pace feels normal. Step two is sections with flags but no edits: you bookmark per the list above, you reach the review screen, you open the flags and you write down what you would have changed, but you change nothing. Then you check. If the changes you would have made are mostly right, your flagging is calibrated. If they are mostly wrong, your flags are feelings.

Step three is the full protocol in sectionals, with changes allowed and a log of every change: question number, the specific doubt, the change, and whether it was right. Step four is the full protocol inside full-length practice tests on Sundays. By the second full test, most candidates have a personal change record that tells them something useful, such as “my Quant edits are right and my Verbal edits are wrong,” and they adjust their caps per section accordingly.

Keep the change log for every practice test. Four weeks of it is the most honest data you will have about whether your second instinct is better than your first.

Section-specific adjustments

Quant edits are the most reliable because the doubts are concrete. A dropped sign or a misread “at least” is checkable in under a minute, and the answer either changes or it does not. Hold four minutes here and expect to use one or two changes on a typical section. Trust Quant edits most.

Verbal is the section where candidates most often make their second answer worse. Critical-reasoning changes are reasonably safe if the doubt was “I never prephrased,” because re-reading the conclusion and prephrasing takes 30 seconds and usually settles it. Reading-comprehension changes are the least reliable, because the doubt is usually “I didn’t go back to the text,” and going back to the text with 90 seconds left under pressure produces a hurried read. Our standing advice is to cap Verbal changes at two and to spend reading-comprehension doubt on going back to the passage during the section, not in the reserve.

Data Insights is where the feature pays best. Multi-part questions are scored all-or-nothing, so a second look that flips one true/false can turn a whole question from wrong to right. The cost is that re-reading a source takes time. Hold four minutes, and prioritize any multi-part set where exactly one part was a guess over a set where two parts were.

What this is worth

We will not quote a point value, because GMAC does not publish one and it depends on how many of your flags were real. What we can say is what the protocol prevents. It prevents the three or four questions per section that candidates lose to a misread constraint they would have caught in 30 seconds. It prevents the time collapse in the last five questions that comes from fighting one problem too long. It prevents the unforced third edit made only because a change was still available. And it prevents the review-screen panic that produces three gut edits in 90 seconds.

Those losses are visible on every Enhanced Score Report we debrief: a timing graph with a spike in the last third and a cluster of misses right where the spike is. The protocol flattens the spike. The rest of your score is content, and content is a different conversation.

What to do this week

1. Take one timed 45-minute section in your weakest area with no flags allowed and the single goal of finishing with four minutes left.
2. Write the seven flag-worthy doubts on an index card and keep it next to you for every timed set this week.
3. Take a second section with flags allowed but no changes; at the review screen write down what you would have changed, then check whether those changes would have been right.
4. Start a change log: question number, the specific doubt, the change, right or wrong. Keep it for every timed section from now on.
5. Check the clock at the two-thirds mark of every section (Q14, Q15, Q13) and practice the best-guess-and-flag move on the next hard question when you are behind.