Every GMAT prep company on the internet sells the same thing: a guaranteed 700. Walk through the marketing pages of the major players and you will see the number repeated like a liturgy — 700, 705, 720, 740. The score is treated as a product feature, as if it could be bundled with a video library and a workbook and shipped overnight.
Now look at the outcome data the test-makers themselves publish. The share of GMAT takers who score 705 or above (the new-scale equivalent of the old 720) is in the low single digits. It has been for years. If even half of the candidates buying premium prep were actually getting what they paid for, that distribution would have shifted by now. It hasn’t.
I want to spend this letter on the gap between those two facts.
Why the industry sells a number it cannot deliver
Prep companies aren’t lying, exactly. They’re responding to incentives. A GMAT candidate is, almost by definition, an anxious customer with a deadline. They have a round-one application six months out, a self-image as a future M7 student, and a very specific number in their head. The companies that win their credit-card information are the ones willing to repeat that number back.
Once you have promised a 700, the rest of the business model writes itself. You package content — because content is cheap to produce and easy to demo. You hire instructors — because parents and partners trust humans more than software. You build a 1,200-page textbook — because shelf weight is a proxy for seriousness. You run search ads against “GMAT 700” and let the funnel do the rest.
None of this is preparation. It is the appearance of preparation, sold to people who have no way to tell the difference until score-release day.
What is missing is measurement
The hard truth is that getting from a 605 to a 705 is not a content problem. By the time a serious candidate sits down to prep, they have usually seen most of the math. They have read sentences in English their entire adult life. What they cannot do is execute reliably, under timing pressure, on the specific sub-skills that the test discriminates on at the top of the curve.
Identifying those sub-skills is not a content task. It is a measurement task. It requires a system that knows, at any moment, what you are actually weak at — weighted by how much that weakness is costing you on the scoring curve, not by how many drills happen to exist in the question bank.
Almost no consumer prep platform does this. The ones that claim to be “adaptive” mostly reshuffle question order or resurface items you got wrong. That is not measurement; it is reinforcement of the wrong loop.
What honest prep would look like
We started Brightroom because we wanted to find out what would happen if a small team built prep from the measurement layer up, and let everything else — content, pricing, marketing — fall out of that decision. Four things follow.
Outcome-aligned pricing. If you charge thousands of dollars for a score, your guarantee has to mean something. A guarantee that lets you keep the money as long as the customer “completed the course” is not a guarantee; it is a compliance checkbox. The real test is whether the company will renew the program when the score doesn’t arrive — without legal gymnastics.
Real measurement. An adaptive system that estimates ability after every response and selects the next item to maximize information. A mastery vector across the actual sub-skills the test rewards. Pacing data treated as a first-class signal, not a footnote.
Adaptive selection, not curated paths. No two candidates should get the same study plan, because no two candidates have the same weaknesses. A “six-week 700 plan” is a marketing artifact, not a preparation strategy.
Transparent guarantee mechanics. Tell the candidate, on the pricing page, exactly what the guarantee promises and what triggers it. If you won’t put the rules in writing, you don’t have a guarantee. You have a sales line.
The Brightroom 715+ Guarantee, in plain language
Here is how ours works. The lift you’re guaranteed depends on the diagnostic score you start from, because a +100 promise to someone starting at 695 is dishonest and a +30 promise to someone starting at 505 is uninteresting.
- Diagnostic 205–505: +110 points, exit at 615.
- Diagnostic 515–555: +80 points, exit at 635.
- Diagnostic 565–605: +60 points, exit at 665.
- Diagnostic 615–645: +50 points, exit at 695.
- Diagnostic 655–675: +40 points, exit at 715.
- Diagnostic 685–745: +30 / +20 points, exits at 735 / 765.
- Ultra: the same banded lift plus an absolute 715+ floor for every Ultra customer. No diagnostic threshold; whichever bar is harder for your situation is the one we’re on the line for.
If we miss after you’ve completed the program — Library complete, three Mocks sat, 600+ practice questions, baseline submitted on time — the next six months are free. New window, same engine, no second invoice. The bands exist because honesty is a function of where you start; we don’t think a single guarantee number can survive contact with the actual distribution of candidates.
Why we’re writing this down
A guarantee on a pricing page is one paragraph. The reasoning behind it is a document. We would rather lose a customer who needs the comfort of a promised number to a competitor than win them by repeating one. The candidates we want are the ones who read this far and recognized, in some private way, that they have been sold the appearance of preparation before.
This is letter No. 01 of a series. We’ll publish openly: how the adaptive engine actually decides what to ask next, how we price, what we’ve gotten wrong, what we think the next decade of standardized testing looks like. We will not edit these to be more flattering after the fact.