GMAT Practice Tests

What separates a practice test from a pile of questions

Most free GMAT material is a question bank with an answer key attached. That has a real use, and we publish one too, but it cannot answer the question you actually care about, which is what you would score if you sat the exam on Saturday. A score is produced by a system, not by a set of questions. The system is a fixed question count, a fixed clock, a section order you commit to before you start, a capped budget of answer changes, and a hard door at the end of every section. Take any one of those away and the number at the end stops being comparable to a real one.

The constraints bite harder than most candidates expect. Twenty-one Quantitative Reasoning questions in 45 minutes leaves a shade under 2 minutes 9 seconds each. Twenty-three Verbal questions in the same 45 minutes leaves 1 minute 57. Twenty Data Insights questions leaves 2 minutes 15, and several Data Insights formats spend a good part of that budget on reading source material before you can begin reasoning at all. A practice test that quietly hands you more time, or trims the question count, trains a pace that will not survive contact with the real interface.

• 45 minutes per section, three sections, 135 minutes of scored testing.
• 21 Quantitative Reasoning questions, 23 Verbal Reasoning questions, 20 Data Insights questions.
• A section order chosen from six permutations before the first section begins.
• One optional 10-minute break, offered after the first section and again after the second.
• Three answer edits per section, spent in the Review Center and nowhere else.
• No return to a section once you have ended its review.

TestLumo reproduces every one of those rules, including the awkward ones. The clock does not stop. The review screen opens only if you finish the section with time left, and it disappears entirely if the clock runs out first, which is exactly what happens on the real exam. Bookmarks carry through to that screen. Changing an answer raises a confirmation dialogue and decrements a counter that starts at three and does not reset. None of this is decoration. It is the part of the exam almost nobody rehearses, and it is the part that quietly costs points.

The three sections, and what is actually in them

Sixty-four scored questions in total, weighted evenly by time rather than by count. That is worth internalising, because it means the three sections contribute comparable amounts to your total despite Data Insights carrying much the heaviest reading load per item. Each section is reported on a 60 to 90 scale and the three combine into a total on the 205 to 805 scale. There is no essay, no separately timed Integrated Reasoning section, and no sub-score beyond those three numbers and their percentiles.

Quantitative Reasoning

Arithmetic and algebra, and nothing else. Geometry has been removed from the current exam, so circles, triangles, coordinate planes and solids are no longer tested and no longer worth revising. Data Sufficiency has also left this section and now sits inside Data Insights, which means all 21 questions are straightforward Problem Solving items with five choices each. There is no calculator, so arithmetic fluency is a scored skill here rather than a convenience.

The concept mix is narrower than most study plans assume. Across the 797 question Official Guide corpus we measured, Applied Problems is comfortably the largest quantitative bucket at 39 items, ahead of Statistics at 22 and Percents at 20. Properties of Numbers accounts for 15. Everything else, including inequalities, combinatorics, probability and estimation, sits in single figures. Our GMAT quantitative practice test reproduces that distribution rather than spreading questions evenly across a textbook contents page, because a proportional test tells you where you are actually losing marks.

Verbal Reasoning

Two formats: Critical Reasoning and Reading Comprehension. Sentence Correction was removed with the Focus Edition, so idiom lists, parallelism drills and modifier rules no longer earn anything. What remains is argument analysis and close reading, which are harder to cram but easier to improve with structured practice, because both reward a repeatable procedure rather than recall.

Critical Reasoning items are remarkably consistent in size. In the measured corpus, half of all official Critical Reasoning items run between 79 and 95 words including the question line, with a median of 87 and a full range of 55 to 139. That tightness is a gift, because it lets you budget reading time almost mechanically. Reading Comprehension passages behave differently: they cluster into two distinct lengths, a short form around 210 to 240 words and a long form around 350 to 380, each carrying three or four questions. Our GMAT verbal practice test is built to both distributions rather than to the average between them.

Data Insights

The section most candidates under-prepare, and the only one with an on-screen calculator. Five formats appear here: Data Sufficiency, Table Analysis, Graphics Interpretation, Multi-Source Reasoning and Two-Part Analysis. Three of them use a split screen with source material on the left and the question on the right, and two use response widgets that are not five radio buttons at all, so the interface itself is something you have to practise before exam day.

Data Insights is where the mathematics and the reading meet. The official score report splits it into Math Related and Non-Math Related content, and the measured corpus shows Applied Problems as the largest bucket at 39 items, followed by inference and application work. Two-Part Analysis carries by far the longest prompts on the exam, a median of 146 words against 36 for a Problem Solving stem, so pacing inside this section is genuinely uneven in a way the other two are not. Our GMAT data insights practice test keeps all five formats in proportion, which is the only way to rehearse that unevenness honestly.

Full length tests, section tests, and when to use each

A full length practice test is a measurement instrument. It costs 135 minutes plus a break, it is tiring, and its output is a score you can trust. A section test costs 45 minutes, produces a section score with the same format-level breakdown, and is the right tool for almost every day that is not a checkpoint. The common mistake is treating them as interchangeable and sitting full length tests weekly because they feel productive. They are not productive. They are diagnostic. The work happens in between them.

A reasonable pattern is one full length test at the very start to establish a baseline, one every two to three weeks after that, and section tests filling the gaps. If you are eight weeks out, that is three or four full sittings and a couple of dozen section papers. Every full length test should be followed by two or three days of targeted work driven by its report, and only then by the next measurement. Sitting two full length tests in a single weekend tells you almost nothing the first one did not already say.

Section tests also let you attack a weakness without paying the fatigue tax. If Data Insights is costing you ten points, you can sit four Data Insights papers in the time one full length test takes, and each produces a format-level breakdown of where the loss actually sits. That is not possible inside a full test, where 20 Data Insights questions are spread across five formats and no single format contributes enough items to be meaningful on its own.

How the questions are built

We measured 797 questions from the current Official Guide and turned the result into an authoring specification. Not the questions themselves, which are copyright and which we do not reproduce, but their shape: stem lengths by format and by difficulty band, answer choice lengths, the spread between the longest and shortest choice, how often choice sets share an opening word, and how concepts are distributed across a section. Every question we write is checked against those bands before it is allowed into the bank.

• Problem Solving stems: median 36 words, half of all items between 20 and 52. Stem length is flat across difficulty, so a long quant question is a scenario, not a hard one.
• Critical Reasoning: median 87 words, half of all items between 79 and 95. The tightest distribution in the entire corpus.
• Reading Comprehension passages: two clusters, roughly 210 to 240 words and 350 to 380, with three or four questions each. The exam caps a passage at four.
• Data Sufficiency: median stem 34 words, statements a median of 13 words each and close to symmetric. Stems grow with difficulty, from a median of 28 at easy to 48 at hard.
• Two-Part Analysis: median 146 words, roughly four times a Problem Solving stem.
• Choice-length spread held inside the measured band, because a conspicuously long correct answer is the oldest tell in the format.

Two authoring rules follow from the data and matter more than any single number. First, every distractor has to be traceable to a specific candidate error, because the official explanations never dismiss a wrong choice as merely wrong; each one names the misstep that produces it. If we cannot name the error a distractor catches, it is filler, and a strong candidate eliminates it on sight. Second, we randomise correct answer positions to a flat 20 per cent each. The published corpus leans towards D at almost 24 per cent against A at under 14, which is an artefact of how the book was assembled rather than a property of the exam, and a bank that inherits that lean is exploitable.

Data Sufficiency is the one deliberate exception to the flat key rule. Its natural distribution genuinely favours the both-statements-needed and still-insufficient cases, because those are the interesting ones to test, so we keep it near its observed shape rather than forcing it flat. Difficulty is also proportioned deliberately: the Official Guide splits its practice set into near-equal thirds of easy, medium and hard items, and our bank matches that, so a paper is not quietly easier or harder than the exam it stands in for.

What a practice score does and does not tell you

Total scores move in steps of ten, which is why a 645 exists and a 647 does not. The three section scores are not simply added; they are combined and mapped onto the total scale, so equal improvements in different sections do not always shift the total by the same amount. Our estimator produces a total on that scale from your section performance, and it is an estimate. It is calibrated to be honest about difficulty and format mix rather than flattering, which occasionally means a practice score that reads lower than the number a free test elsewhere handed you.

What a practice score is genuinely good for is direction and stability. One test tells you roughly where you are. Three tests across six weeks tell you whether the plan is working, and that trend is far more informative than any individual figure. Treat a single result as a data point with error bars around it, not as a prediction, and never make an application decision on the strength of one sitting.

A sequence that works

1. Sit one full length practice test cold, before any revision, and accept whatever it says. A baseline taken after two weeks of study is not a baseline.
2. Read the report by section and by format, never by total. The total contains nothing actionable.
3. Pick the two weakest format-level buckets and work only on those for a week, using section tests to check.
4. Drill timing separately from accuracy. They are different problems and mixing them hides both.
5. Re-measure with a full length test every two to three weeks, keeping the same section order you intend to use on exam day.
6. In the final fortnight, stop adding new material and rehearse the interface: section order, break placement, edit budget, review screen.

The step people skip is the fourth. Accuracy problems and timing problems look identical on a score report but need opposite treatments. If you are getting questions wrong because you rushed, more practice at speed makes it worse. If you are getting them wrong because a concept is missing, untimed practice fixes nothing about the clock. Separate them by sitting one section untimed, comparing that accuracy against your timed result, and treating the gap between them as the price you are currently paying for pace.

Sitting a practice test so the score is worth having

• Sit it at the hour your real exam is booked for. Cognitive performance is not flat across a day.
• One screen, no phone, no second tab. A single lookup invalidates the section.
• Use scratch paper and a pen. On the day you get a laminated booklet and a marker, and it is slower than paper, so build in the friction.
• Choose your section order once and keep it for every test. Comparability depends on it.
• Take the break, and take it in the same place every time. Ten minutes is enough to eat something and not enough to lose focus.
• Do not pause, do not restart, and do not look anything up mid-section. Finish the bad section; the recovery is itself the skill being tested.

That last one is worth defending. Almost everybody has one section that goes badly, and the difference between a good total and a poor one is usually what happens in the section immediately after it. Candidates who have only ever practised sections in isolation have never rehearsed carrying on after a disaster. That is a specific, trainable skill, and a full length test is the only place you can train it.

Reading the report afterwards

Every TestLumo test produces a report built on the same taxonomy the official score card uses, which is deliberate: using the official labels makes our analytics directly comparable to a real score report. You get a score for each section and a total on the 205 to 805 scale, then a per-question table giving the answer you chose, the marked answer, the time you spent and a worked solution — for every question, including the ones you got right. What you will not find is a percentile. We have no calibration sample to compute one honestly, and an invented percentile is worse than none.

The single most useful panel is the time-against-outcome scatter. It plots every question by how long you spent and whether you got it right, against a two-minute reference line. Four clusters matter: fast and right, which is fine; slow and right, which is where your time is leaking; fast and wrong, which is carelessness or a guess; and slow and wrong, which is the expensive one, because you paid full price for a question and got nothing back. A candidate with six slow-and-wrong items in a section has just found their entire timing problem in one chart.

• Quantitative Reasoning content domain: Arithmetic and Algebra.
• Quantitative Reasoning question type: Pure Contexts and Real Contexts, roughly half each.
• Verbal Reasoning question type: Critical Reasoning and Reading Comprehension.
• Data Insights question type: Data Sufficiency, Graphs and Tables, Two-part analysis, Multi-source reasoning.
• Verbal fundamental skills: Analysis/Critique, Plan/Construct, Identify Stated Idea, Identify Inferred Idea.
• Quantitative fundamental skills: Value/Order/Factors, Equal/Unequal/ALG, Rates/Ratios/Percent, Counting/Sets/Series/Prob/Stats.

The report does not currently record which of your three answer changes per section you spent, or what you changed each answer to, and that is a gap worth naming: which way your changes tend to go is one of the few things about your own testing that nobody else can tell you. Until it is there, note it yourself during review. If your changes are mostly right to wrong, spend the edit budget only on questions you flagged as genuinely unresolved during the section, not on questions that merely feel uncertain in hindsight.

Four ways people waste practice tests

Sitting them too often

A full length test every few days produces a flat line and a tired candidate. There is no mechanism by which your score improves between Tuesday and Friday if the only thing you did in between was sit another test. Measurement is not training, and treating it as training is the most common way an eight week plan produces no movement at all.

Reviewing only the wrong answers

A question you guessed correctly is a question you will get wrong next time. Review anything you were not confident about, and use the response-time column to find those items: something you answered correctly in three and a half minutes was not really a win, it was a near miss that happened to land.

Practising with the clock off

Untimed practice has a place, but it is a different exercise with a different purpose. If most of your preparation is untimed, your first timed section on exam day is effectively your first honest attempt at the actual task, and the pacing shock is worth a substantial number of points.

Changing section order between tests

Six permutations are available and any of them is defensible, but switching between tests destroys the comparability of your own results. Pick the order that puts your strongest section first if you need early confidence, or your weakest first if you fade badly across a long sitting, and then stop changing it.

Where to start

If you have never sat one, start with the full length practice test and give it the full 135 minutes. If you already have a baseline and know roughly where the damage is, go straight to the quantitative practice test, the verbal practice test or the data insights practice test and work the weakness directly. Everything is free, nothing requires a card, and the reports are identical whichever route you take.