The story behind the score.

Iwas 26, working full time at a consulting firm in Yerevan, and the first time I sat for the GMAT I scored 565. I'd been studying for three months, doing what every guide tells you to do — work through chapters, do practice sets, take mocks, repeat. I was averaging two hours a day. I felt like I was getting better. I wasn't.

The 565 was a body blow. I'd been targeting M7 schools. Median GMAT at HBS / Stanford / Wharton was 730+ at the time. The gap between where I was and where I needed to be was roughly 170 points — and the prep cycle I'd just completed had moved me 15 points off my untested baseline. At that pace, I was on track to hit 730 around 2032.

I did what everyone does. I panicked and bought a more expensive course. Watched more videos. Did more questions. After eight more weeks of this, my mock scores had moved another 10 points. The course was thorough. The instructors were competent. None of it was the problem.

The problem — and this took me three months to actually understand — was that I was practising without reviewing. I'd get a question wrong, read the explanation, nod, and move on. Three weeks later, I'd miss the same question type again. I'd read the same explanation again. I'd nod again. The content went in and out without ever leaving a mark on what I'd do differently next time.

The shift came in month four. I built a spreadsheet. Six columns: question source, section, what I got wrong, why, fix, retest. Six tags for the “why” column: Conceptual, Careless, Time pressure, Misread, Strategy, Other. One row per missed question. No exceptions.

For the first two weeks I just logged. No drilling, no re-attempts, no plan. Just every miss into the spreadsheet, with five fields filled in honestly.

At week three I sorted the log. The pattern was almost embarrassing. Out of 140 missed questions, more than half were three things: inequality sign-flips on Algebra, CR Assumption questions where I'd eliminate the right answer because it “felt too broad,” and DI Table-Analysis questions where I tried to compute exact values instead of comparisons. Three patterns. Across hundreds of questions. The content was never the problem.

I built a study loop around the log. Drill 20 questions on a target topic, log every miss with the five fields, end the week reviewing the log and doing a 30-question targeted set on whatever pattern dominated. Re-test flagged questions a week later. From 565 in October to 645 in late January. 695 in March. 735 in May. 735. 100th percentile. Eight months from the worst day of my prep cycle.

The slope wasn't linear — most of the climb happened once the log had eight to ten weeks of data and the patterns were visible. Below that threshold, you don't have enough signal to act on. The discipline was in logging through the noise until the signal showed up.

Zakarian GMAT is the spreadsheet, rebuilt as a real product. The error log is built in with the same six tags. The analytics surface the patterns automatically — you don't have to remember to sort. The adaptive study plan re-prioritises every week based on the log's latest data. The spaced review queue resurfaces past misses on the schedule that actually makes them stick. Coaching is available for a second set of eyes on your patterns; the platform alone is the loop you can run by yourself.