What This Episode Covers
How to review practice questions gets debated endlessly online, and Isaac thinks it shouldn't be. The process he teaches is straightforward, and it's probably a little different from the review habits you built studying for other exams. Some of it might even feel like it's slowing you down, especially at first. That's common and normal, and the episode addresses that feeling directly: if what you're doing is working, keep doing it. If you're putting in the time and energy and still not seeing the gains you want, the problem probably isn't effort. It's process.
The episode walks through review section by section, with worked demonstrations for each. For quant questions, the process starts the moment you check the answer: web search the question, collect alternate approaches from forums or an AI, compare them against yours, and write out by hand any approach that beats yours. For verbal, the process looks different: whether you got the question right or wrong, you write out why the right answer is right and why each wrong answer is wrong, in your own words. For Data Insights, data sufficiency follows the quant process, and verbal-style questions follow the verbal process.
Two ideas run underneath all of it. The purpose of a practice question is to learn from it, not to complete it, which is why questions you answer correctly get the same review as questions you miss. And the learning science behind the method, reflecting on your own reasoning, is what moves your practice gains from the questions you've already seen to the questions you'll meet for the first time on test day.
The Quant Review Process
After you attempt a quant question and check the answer, the review has three steps.
Web search the question. Any search you're comfortable with works. A classic web search usually turns up forum threads with several human explanations. If you'd rather use AI, that's viable too, though check your question provider's policy first: some providers have terms about pasting their questions into AI tools, and Isaac is careful to say he's not giving legal advice on that point. His take is practical: many students are going to use AI regardless, so it's worth learning to use it well.
Collect alternate approaches. Once you've read an explanation (or asked an AI for one), go further: ask what other ways exist to approach the question, or scroll the forum thread and glance at every method posted. Compare each one against the approach you used. The obvious reason is that if a better way exists, you want to know it. The less obvious reason comes from learning science: comparing approaches builds the reflection step that makes knowledge stick for unseen questions later, even when no better solution turns up.
Write out anything better, by hand. If one of the alternate approaches beats yours, re-copy the full solution yourself on paper. Something different happens in your brain when your hands are involved, and it doubles as practice for the scratch-work habits you'll need on exam day. You're slowing the test down to do an out-of-time analysis of one question, which can feel unproductive in the moment. Long term, it's the fastest path Isaac knows to the results you want, and if it weren't, he wouldn't recommend it.
One logistical tip from the episode: time it. Use a count-up timer and compare how long the full alternate-approaches pass takes against glancing at one explanation and moving on. Unless you fall down a rabbit hole on a topic with lots of variety, the extra time is usually small, and it's buying skill diversity in the section where your approaches drive your results.
Reviewing Questions You Got Right
It's intuitive to think more reps equal more score, and for a small percentage of natural test takers, they do. If that's you, Isaac won't argue with your results; he'll be the first to say keep doing what you're doing. His own story was the opposite. He struggled heavily with this exam, tried the brute force approach, watched his score go down, and nearly gave up more than once. The process in this episode is what he wishes he'd had from the start.
If you answer a question correctly and move on without reflecting, you've collected a confidence boost and not much else. Skills are what drive your score on test day, not the number of questions you complete. And the learning science here is the part worth knowing: studies on metacognition (thinking about how you thought through something) show little difference between groups in the moment. The gap opens up later. Test the same groups days or weeks afterward, and the reflection group significantly outperforms. Since the only performance you care about is on a date weeks away from your practice sessions, the delayed effect is the one you want working for you.
You probably won't spend as long reviewing questions you got right as questions you missed, and that's fine. The point is to at least go through the process.
The Two Lists for Questions You Missed
Wrong answers get logged, but not in one pile. The traditional single-list error log tends to become a garbage can of wrong answers that rarely gets revisited and doesn't produce much. Isaac's system separates misses into two categories.
Category A: you knew how to do it and missed it anyway. You picked the right strategy, executed it well, and then added five plus five and got eleven, or solved for X when the question asked for Y. For each Category A miss, name the mistake in your own words ("I solved for the wrong quantity") and write the new behavior that prevents it next time ("circle what the question is asking for before I solve"). New entry, every time the mistake happens. The point is awareness: once you know your top miss reasons, your brain starts flagging those situations on its own, and you slow down exactly where you need to.
Category B: you didn't know how to do it, or not well enough. If you could get it right in ten minutes but the exam allows about two, that belongs here too, along with any question where you had no strategy at all. For each Category B miss, record where to find the question so you can re-solve it, and write a cue for what you want to think next time you see a question like it. It works like a flashcard: the problem is the front, the strategy is the back. Tracking your history with each question (dates, right or wrong, time per attempt) is optional but valuable; you can watch your recognition speed improve over weeks.
Then put the Category B list to work: re-solve five to ten questions from it at the start of every study session, more if you have time. It's a warm-up for your brain before the heavy part of the workout. Execute on it for a while and you should notice new questions calling up old strategies faster.
Any format works for either list. A simple spreadsheet is fine, pen and paper is fine if you want less screen time, and the elaborate custom setups some students build are fine too. Isaac's a fan of the nerdy builds; he just doesn't want anyone thinking the tool is the point. The logging is one part of building skills. Putting the lists into practice is the part that moves your score.
Using AI to Review (Without Wasting Time)
The episode keeps AI in its place: useful, optional, and easy to waste time with if you use it carelessly. Two tips do most of the work.
First, turn on thinking or extended thinking mode if your provider offers it. Second, tell the AI the correct answer before you ask it to explain anything. Don't test the AI on the question first. It may get the question wrong, and then you can burn a lot of time being persuaded of an answer you know is wrong.
Beyond that, an AI "project" (a folder where your chats all draw on the same instructions and materials) is a worthwhile upgrade if your provider supports it. There's a full blog post on that setup linked below, plus a whole episode on using AI in GMAT® prep. Isaac's framing on whether you need any of it: you don't. You can get a great score without AI, and at the current state of the technology it isn't necessarily a large competitive advantage in GMAT® prep anyway. If you're going to use it, use it well.
Reviewing Verbal Questions
Verbal review has a different shape, and it's simpler.
Whether you got the question right or wrong, after checking the answer you write two things: why the right answer is right, in your own words and your own logic, and why each wrong answer is wrong. That's the quick start.
The demonstration uses a reading comprehension passage from mba.com's sample questions about analog versus digital recording. On a question where the answer is (B), the review asks what makes (B) right, then takes each wrong choice in turn. One wrong choice describes something the passage never mentions, so the log entry is one line: not in the passage, therefore not the answer. Another wrong choice is supported by the passage but responds to a question the test isn't asking, so that entry reads: this doesn't answer the question I'm being asked. Both are plain-language diagnoses that pay off the next time you see a similar trap.
Two things to skip for verbal. Don't write out alternate approaches; that's a quant-only step, and verbal questions rarely have multiple valid paths. And be careful with forum threads on verbal questions: there are usually one or two good explanations buried under a lot of reasoning that won't help you, which makes them a worse use of time than quant threads. If an explanation doesn't land, asking an AI to explain it a different way does have value. Just don't ask it for alternate approaches. For deeper verbal help, Isaac's suggestion is a skilled professional or a highly rated, data-backed course, unless verbal comes to you naturally, in which case: ride the wave.
Critical reasoning follows exactly the same process as reading comprehension. Nothing new to learn.
If you study on a platform that forbids searching its questions (some providers protect proprietary content), use the platform's own explanations of why right answers are right and why wrong answers are wrong. Then do the written step anyway. The writing is the metacognition, and the metacognition is the part that transfers.
Reviewing Data Insights Questions
Data sufficiency gets the exact quant process from above, with no differences.
For the rest of the section, DI questions blend math and verbal in different proportions: some are pure math, some are reading comprehension or critical reasoning, and many mix the two. The episode's rule for handling that variety is to judge which kind the question leans toward, then use the matching review process. A two-part analysis built on rates formulas gets the quant treatment. A two-part analysis built on a long passage gets the verbal treatment, with one bonus: because a two-part analysis has two correct answers in one question, a full review means writing out why each correct answer is right and why each wrong option is wrong, for both parts. Twice the learning from a single question, which is the glass-half-full way to look at one of the section's heavier question types.
Key Takeaways
- Review right answers too. The purpose of a practice question is to learn from it, not to complete it. Skills drive your test-day score, not question count.
- For quant: web search, compare, write. After checking the answer, find alternate approaches (forums or AI), compare them to yours, and hand-copy any solution that beats yours.
- Tell the AI the answer first. If you use AI for review, turn on extended thinking and give it the correct answer upfront, so it spends its effort explaining instead of persuading.
- Separate misses into two lists. Category A: knew it, missed it anyway. Name the mistake, write the fix, watch for patterns. Category B: didn't know how to do it. Log the source, write a next-time cue, re-solve it.
- Re-solve five to ten Category B questions per study session. Treat it as a warm-up. Recognition speed should improve over weeks.
- Verbal review is written, not searched. Whether right or wrong: why the right answer is right and why each wrong answer is wrong, in your own words. No alternate approaches for verbal.
- DI inherits both processes. Data sufficiency follows the quant process. Verbal-leaning DI follows the verbal process. Use your judgment on the blend.
- Adopt it gradually. Add one or two techniques per study session, then let the data (accuracy, speed, practice test results) tell you whether it's working.