GMAT Scoring Explained

How Is the GMAT Scored?

The GMAT consists of 64 questions across three 45-minute sections: 21 Quantitative Reasoning questions, 23 Verbal Reasoning questions, and 20 Data Insights questions. Each section produces a score on a 60-90 scale, and the three scores combine, with equal weight, to produce your 205-805 total. To see how section scores translate into totals, try our GMAT score calculator.

Two things about that structure tend to catch students off guard. First, equal weight means equal weight: one point of Data Insights improvement moves your total exactly as much as one point on Quant, so prep plans that treat DI as a footnote are optimizing for an exam that no longer exists. Second, there’s no essay anywhere in the scoring. The old Analytical Writing Assessment was removed in 2024.

In case you’re wondering about the strange-looking score scale: totals run in 10-point increments and always end in 5, a deliberate GMAC design choice so that current scores can’t be confused with scores from the previous GMAT (the 10th Edition), which ended in 0. The full story about the GMAT scale is in: What Is the GMAT Scored Out Of?

Inside the GMAT Scoring Algorithm

How exactly does the GMAT scoring algorithm work? Why do some questions feel harder than they’re labeled? We sat down with Jaime Malatesta, Senior Psychometrician and Research Manager at GMAC, to cut through the myths.

The first thing to understand: the GMAT isn’t testing whether you memorized obscure math formulas or tricky grammar rules. The exam is designed to predict your likelihood of success in business school. As Jaime explains, psychometricians measure “things you can’t see or touch with a ruler,” in this case, your readiness to succeed in a rigorous MBA, PhD, or other business-related graduate program.

The GMAT uses an adaptive scoring algorithm based on Item Response Theory (IRT). In plain English:

• Your score is based on the difficulty level of the questions you answer correctly and incorrectly, not just the number of questions you answer correctly.
• Each question has quantifiable characteristics (such as difficulty), calculated from the performance data of hundreds or thousands of real test-takers.
• The algorithm updates after every response to estimate your ability and deliver the next question, tailored to your skill level.

So the test isn’t counting “how many” you got right; it’s assessing which questions you answered correctly and what that reveals about your ability. 

The difference is dramatic in practice: miss eight relatively easy Quant questions and your section score might land in the 60s or low 70s, but miss eight much harder questions and you could still walk away with an 80Q or higher. If you perform well, the questions get harder, and your ceiling rises with them. 

That’s also why careless errors on easy questions are the most costly mistakes on the test: the algorithm reads them as a signal about your ability, and it will give you easier questions that limit your ability to achieve a strong score.

Why “Easy” and “Hard” Don’t Always Feel Right

Many test-takers say: “That problem felt easy, but it was labeled as hard.” Here’s why. Difficulty on the GMAT is 100% data-driven, not subjective. Every new question begins its life as an experimental item, tested with real candidates but not scored, and is then statistically calibrated based on how those candidates actually performed. GMAC staff never sit around debating whether a question feels easy or hard, so your perception of difficulty may not match the statistical classification you see on a score report.

Experimental Questions: What You Should Know

Yes, there are experimental questions on every GMAT section. They don’t count toward your score, they appear in random positions, and they exist so GMAC can collect the data needed to calibrate new items. How many are there? GMAC won’t say, and for good reason: as a test-taker, it doesn’t help to waste energy trying to identify experimental questions. Treat every question as if it counts.

Does Luck Play a Role in Your GMAT Score?

Not really. Every test is carefully balanced to cover the full content domain, and security protocols keep overused questions from appearing too often. Scores can fluctuate slightly, about ±2 points per section and ±24 points on the total, which is normal measurement error. You might see small variations from day to day, but your score is a stable, reliable reflection of your ability level. (Keep that ±24 in mind when you’re deciding whether a retake is worth it; we’ll say more about retakes below.)

Time Pressure Is Part of the Design

Why isn’t there unlimited time? Here are the main reasons, per GMAC: fairness and practicality; predictable schedules for test-takers and test centers; and real-world alignment. Business leaders don’t have infinite time to analyze data, and the GMAT measures whether you can reason under realistic constraints. Everyone would appreciate more time, but the exam is designed so that most candidates have enough to finish.

How Specific Question Types Are Scored

• Multi-part Data Insights questions (e.g., three sub-questions on one graph) are all-or-nothing. No partial credit.
• Reading Comprehension passages come as bundled sets: once you get a passage, you’ll see all of its attached questions, regardless of how you answered the first one. Expect 3-4 passages, short or long.
• Multi-Source Reasoning typically appears in 1-2 sets per test.
• Unanswered questions are penalized heavily. An incorrect answer typically costs you far less than a blank one, so never let the clock run out with questions unanswered.

These variations don’t change the overall difficulty balance; the algorithm ensures fairness across all candidates.

Practice Tests vs. the Real GMAT

GMAC’s official practice exams on mba.com are the closest you can get to the real thing: they use the same scoring algorithm as the live test, and their questions are retired GMAT items with valid statistics. The only difference is that practice tests contain no experimental questions, meaning every item counts, which actually gives you more diagnostic information while you prepare.

There’s a reason official material is in a class of its own. When Dr. Lawrence Rudner was GMAC’s Chief Psychometrician, he put the development cost at $1,500 to $2,500 per question, in 2007. Adjusted for nearly two decades of inflation, that’s comfortably past $3,000 per question today, against the $5 to $30 a typical prep company might spend on a knockoff. Every official question survives a seven-step gauntlet: expert drafting, quality-rubric checks, editorial review, an international fairness panel, live experimental pretesting, and statistical calibration.

The full economics are here, and they’re why we tell every student: when your score matters, practice on the real thing.

GMAT Percentiles: The Number Behind the Number

A GMAT score without its percentile is only half the story. Your percentile shows the percentage of test-takers you outperformed, and on the current exam, percentiles matter heavily because the 205-805 scale was deliberately rebuilt. The old 200-800 scale’s scores had drifted upward into an uneven distribution over two decades, so GMAC reset the curve in 2024, and schools now read percentile rankings, not how a score “sounds.” Each section carries its own percentile too, alongside the total.

Percentiles are also the answer to one of the most common complaints we hear: that Data Insights scores look lower than Quant or Verbal. You can’t compare raw section scores across sections. Per GMAC, an 80 on Data Insights currently sits around the 84th percentile, while an 80 on Verbal sits around the 57th. Same number, wildly different meaning. 

Use percentiles, not raw section scores, to interpret your performance, and if DI is the section dragging your total down, our Data Insights guides cover every question type.

The scale rebuild is also why you can’t compare current scores to old ones directly. GMAC is blunt about this: comparing old and new scores point-for-point is “not appropriate, accurate, or meaningful.” Instead, GMAC publishes an official concordance (most recently updated July 2025) that matches scores by percentile. 

A few anchor points: a 645 today corresponds to the old 700 (roughly 87th percentile), a 725 corresponds to the old 760 (99th percentile), and a 565 corresponds to the old 600 (right around the middle of the distribution). If you’re comparing your score to an old one, or to a school’s published pre-2024 class average,  our conversion guide walks through the entire table.

What Counts as a Good GMAT Score?

A good GMAT score is one that’s competitive for the programs you’re targeting. There’s no universal cutoff. A 645 puts you around the 87th percentile and a 655 around the 90th, elite territory at most programs, while a 555 sits near the middle of the distribution, and any of those numbers might be “good” depending on where you’re applying. 

The right benchmark is your target schools’ published medians and admitted-student ranges: meet or beat them and your score is doing its job, even if the same number would fall short somewhere more selective. 

One warning from years of tutoring: chasing a “perfect” score you don’t need is one of the most common ways applicants lose months of valuable study time.

For perspective while you chase your number, it’s worth knowing what the GMAT actually predicts. The research says your score predicts first-year business school grades, correlations run around 0.3 to 0.55 across studies, which is exactly why admissions committees use it. But the same research finds it predicts almost nothing about employability, leadership, or long-term career success; one business school’s own employability study found the GMAT “close to irrelevant” and stopped relying on it. 

And when we lined up the 20 most talented students we’ve ever taught, only half had scored 645 or better, five were stuck around 545 despite herculean effort, and four of those five are now at top-ten business schools. We help people raise scores for a living, and we still think the number deserves less reverence than it gets.

Your Scores Are Private Until You Say Otherwise

Here’s the part of GMAT scoring that removes most of the fear: schools never see your scores, or even that you took the exam, unless you choose to send them. There’s nothing to “cancel”; a disappointing score simply stays private while you prepare to retake the exam (though you’ll have to wait at least 16 calendar days between attempts). The one decision you can’t take back: once you send a score to a program, it can’t be un-sent.

Should you ever send a lower score on purpose? Sometimes, yes. One of our favorite students climbed from a 580 to a 710 (on the old scale) in about two months, and deliberately sent all three scores, because the visible climb backed up the work-ethic story in the rest of his application. He landed at an Ivy League program. With superscoring now highlighting your best sections automatically, sending your history is less risky than it used to be, and occasionally it truly helps your applications.

There is, however, a different and scarier kind of cancellation: GMAC itself can cancel your score over a test-security violation, even an unintentional one. Honest test-takers stumble into this more often than you’d think, via touching a phone during a break, chatting in the hallway, or even a banned item as innocent as chapstick. The reassuring numbers: about 94% of cancelled scores are never seen by schools at all, and every appeal is reviewed by a real person. But the best move is prevention, and the guide linked above (built from our conversations with GMAC’s test-security team) covers every common tripwire.

Superscoring: Your Best Sections, Officially Combined

The GMAT now offers a superscore: your best score from each section across multiple attempts, combined into a single total that appears on your Official Score Report, rolling out in 2026.

Here’s how the math works. Suppose you take the exam three times and score 525 each time, first with a strong Quant day (80Q/74V/74DI), then a strong Verbal day (74Q/80V/74DI), then a strong DI day (74Q/74V/80DI). Your superscore combines the best of each: 80/80/80, a 605. Three identical totals become a meaningfully higher number, and because sections weigh equally, even a 3-point improvement in one section can move your superscore by 20 points. When you send a score, schools see three things: the individual attempt, the superscore, and which test date produced each best section.

Two eligibility notes: superscoring applies automatically to all current-edition attempts, and old 10th Edition scores (the ones ending in 0) don’t qualify. Our superscore guide covers what schools see and don’t, how they’re likely to use it, and what, if anything, you should change about your retake strategy. Spoiler: probably nothing drastic.

Your GMAT Score Report: Timeline and What’s In It

You’ll see your unofficial total the moment you finish the exam. It’s your real score, just not yet the version schools can access, because it needs to be verified first. The official report typically posts to your mba.com account within 3-5 business days, though it occasionally takes up to 20.

That report contains far more than the number: question-level performance, timing data, and content-domain breakdowns. Some of it will help you diagnose issues, and some of it is noise. The “Performance by Fundamental Skills” charts, for instance, can rest on as few as three or four questions per category, too small a sample to take literally, while the timing data is gold: bad starts, bad finishes, and wildly uneven pacing are the first patterns we check in our students’ reports, because they usually point to anxiety, warm-up, or stamina issues rather than knowledge gaps. Here’s the full tour of which parts matter.

Two logistics notes to round it out: GMAT scores stay valid for five years from your test date, after which schools can no longer access them, worth checking if you’re testing early or returning to b-school plans after a gap, and the mechanics of sending scores to programs are covered in What Do Schools Actually See On Your GMAT Score Report?

When the Score Isn’t What You Wanted

We recommend beginning with a decision framework: a retake is worth it only if your score is genuinely holding your application back and you can identify clear, actionable ways to improve. GMAC’s own data explains why that second condition matters so much: the average retaker improves by only about 30 points, roughly a quarter of retakers score lower the second time, and normal sitting-to-sitting variation is itself close to 30 points (remember that ±24 measurement error). 

If you’re within about 30 points of your target and you’ve fixed what went wrong, a retake has a reasonable chance; if you need a 100-point jump, you’re planning a study campaign, not a retake.

If your score has stalled across attempts, or seems to drop the more you study, your math, reading, and data skills are probably not getting worse. Something specific is usually going on: inconsistent practice-test data, memorization-heavy studying that quietly lets other sections atrophy, or, most commonly, execution problems such as time management, careless errors, and test anxiety, which gets worse over time if left unaddressed, not better.

And if your score simply isn’t approaching a level that’s competitive for your target programs, you have four practical paths: switch to the GRE or Executive Assessment (the EA uses almost exactly the same question types with less time pressure; the GRE’s different format genuinely suits some people better), rethink your target programs, consider part-time or Executive MBA routes where scores weigh differently, or reconsider whether an MBA is the right destination at all. A low GMAT score has ended far fewer MBA dreams than the internet would have you believe.

Section-by-section improvement lives in our deep guides: Quantitative Reasoning, Verbal Reasoning, and  Data Insights. And for more from inside GMAC, watch our full Straight Talk with GMAC interview series.