CAT Exam Pattern: Sections, Marking, Question Types and Percentiles

Before you plan a single study hour, you need a clear picture of the paper you are training for. The CAT pattern decides how you should practise, how you pace yourself, when you should guess and when you should skip. This guide walks through the current pattern section by section, explains the marking and question types, shows how the paper has changed in recent years using the actual question counts from past papers on this site, and explains normalisation and percentiles in plain words.

CAT at a glance

The Common Admission Test is a computer-based exam conducted by one of the IIMs in rotation. It is usually held in late November, in three slots, and the same pattern applies to every slot.

Details such as eligibility, registration windows, fees and the exact slot timings change every year, so always check the official CAT website for the current notice. The structure above has been stable for several years and is what you should train for.

The three sections

VARC: Verbal Ability and Reading Comprehension

VARC has carried 24 questions in every paper since 2021. Reading comprehension dominates: typically four passages with four questions each, so around two-thirds of the section is RC. The remaining questions are verbal ability, drawn mostly from para jumbles, para summary, odd sentence out and missing sentence (sentence placement) questions. Para jumbles and odd-one-out questions are usually TITA, which matters for guessing strategy (more on that below).

Old-style grammar, vocabulary and fill-in-the-blank questions have not been a regular part of recent CAT papers, though they still appear in other MBA exams. Spend your VARC hours mainly on reading comprehension practice, followed by para jumbles and para summary. The VARC preparation guide covers the approach in detail.

DILR: Data Interpretation and Logical Reasoning

DILR is built from sets: a block of data or a puzzle followed by four to six questions. A typical recent paper has four or five sets. You will not solve all of them in 40 minutes, and you are not expected to. The skill being tested is choosing the right two or three sets to crack fully, so the first few minutes of the section, spent scanning sets and judging which ones open up, are the most important minutes in it.

Common set types include arrangements, games and tournaments, selection and distribution and tables and graphs, but CAT sets often blend types or invent new formats. See the DILR preparation guide for set selection in practice.

QA: Quantitative Ability

QA has carried 22 questions in every paper since 2021. The questions are individual problems, not sets. Arithmetic (percentages, ratios, profit and loss, time-speed-distance, time and work, mixtures) and algebra (equations, inequalities, functions, progressions) usually make up the bulk, with geometry and mensuration next, and number system and modern maths (permutations, probability, sets) filling the rest. The exact mix varies by slot, which is why you cannot skip whole areas. The CAT syllabus guide lists the topics, and the Quant preparation guide covers how to build speed.

Question types: MCQ and TITA

MCQ questions give you four options and carry negative marking: +3 if right, −1 if wrong.

TITA (type in the answer) questions have no options. You type a number, or for a para jumble a sequence such as 3142, using an on-screen keypad. A wrong TITA answer costs nothing. In recent papers, roughly a quarter to a third of the questions in each section are TITA, with QA and DILR usually carrying more of them than VARC.

Two practical consequences follow:

• TITA questions are harder to fake. You cannot back-solve from options or eliminate choices, so you must actually solve the problem. In QA, plugging options into the question is a useful MCQ technique that simply does not exist for TITA.
• TITA guessing is free. If you have worked a TITA question most of the way and have a reasonable answer, enter it. An empty TITA box and a wrong TITA box score the same.

The marking scheme, worked through

Every question is worth 3 marks, regardless of section or difficulty. So a 66-question paper is out of 198 and a 68-question paper is out of 204.

Suppose in QA you attempt 14 questions: 9 MCQs and 5 TITAs. You get 7 MCQs and 4 TITAs right.

• Correct: 11 × 3 = 33
• Wrong MCQs: 2 × (−1) = −2
• Wrong TITA: 1 × 0 = 0
• Section score: 31

Notice that the two wrong MCQs cost you less than one correct answer earns. Negative marking punishes careless attempts, but it does not justify extreme caution.

When does guessing an MCQ make sense?

A pure random guess among four options has an expected value of zero: a one-in-four chance of +3 and a three-in-four chance of −1. So blind guessing neither helps nor hurts on average; it just adds noise. The picture changes once you can eliminate options:

The rule that follows: do not blind-guess MCQs, but an educated guess after eliminating one or two options is worth taking. Track your educated guesses in mocks to check that your elimination is actually reliable.

How the CAT pattern has changed over the years

CAT has been through several redesigns. It was a paper-based test until 2008 and moved to computer-based testing in 2009. From 2015 to 2019 the paper had 100 questions in 180 minutes, with 60 minutes per section. The biggest recent change came in 2020, when the duration was cut to 120 minutes with 40 minutes per section. The number of questions then settled further.

The table below is built from the question counts of the actual CAT papers in our database. Counts are per slot and were the same in all three slots unless noted.

What the numbers tell you:

• VARC and QA have been stable at 24 and 22 questions since 2021. Expect that, but do not be thrown if it moves by one or two.
• DILR has crept up from 20 to 22 questions, which usually means one extra set or a set with more questions in the same 40 minutes. Set selection matters more than ever.
• Time per question is tight: in QA, 40 minutes for 22 questions is under two minutes each, before you account for reading and checking. Nobody is expected to attempt everything.

Because the core design has held steady since 2020, papers from 2020 onwards are the closest thing to a real mock you can get. All of them are on the CAT previous year papers page with full solutions, and you can take them as timed tests in the exam order.

Slots, normalisation and percentiles in plain words

Why normalisation is needed

CAT is held in three slots with three different papers. However carefully they are set, one slot will turn out a little harder or easier than another. If raw marks were compared directly, a candidate in a harder slot would be at a disadvantage through no fault of their own.

Normalisation fixes this by comparing you with the people in your own slot first. Roughly speaking, if you scored better than 90% of the candidates in your slot, your normalised (scaled) score is set to match what a candidate at the same position in another slot achieved. The assumption is that a large, random group of candidates in each slot has the same spread of ability, so the same relative position should mean the same score. The exact method is published by the conducting IIM, so check the official notice if you want the formula; for preparation, what matters is that your rank within your slot drives your result, not your raw marks.

What a percentile actually means

A percentile is not a percentage of marks. It tells you what share of test-takers scored below you. If you are at the 95th percentile, 95% of candidates scored lower than you. If two lakh people take the test, the 99th percentile means you are in roughly the top 2,000.

CAT reports an overall percentile and a percentile for each section. Many institutes use sectional percentiles as minimum cutoffs alongside the overall one, so a very high overall score with one weak section can still fall short. The cutoffs themselves vary by institute and year; check each institute's admissions page.

Why raw scores can look low

In a typical CAT, a raw score that would look modest in a school exam can translate into a high percentile, because the paper is designed so that very few people finish it. This is useful to remember during mocks: your score as a percentage of the maximum is far less important than how you compare with other serious candidates.

What the pattern means for your strategy

Train for 40-minute blocks in the fixed order

Because you cannot move time between sections, each section is effectively its own 40-minute exam. You cannot bank time from a smooth VARC to rescue a tough DILR. Practise sectional tests of exactly 40 minutes and full tests in the VARC → DILR → QA order, so that you get used to starting DILR right after reading four dense passages and starting QA when you are already tired.

Selection beats speed

With under two minutes per question and some questions designed to waste time, the best scorers are not the fastest solvers but the best selectors. In QA, aim to read every question once and solve the ones that open up quickly before returning to the rest. In DILR, spend the first three to five minutes scanning all sets before committing to one. In VARC, choose your passages if the time does not allow all four.

Use the TITA rule and the guessing rule

Always enter an answer in a TITA question you have genuinely worked on. Never blind-guess an MCQ, but take educated guesses once you have eliminated one or two options.

Do not neglect your weakest section

Sectional percentiles and sectional cutoffs mean you cannot carry a weak section on the back of a strong one. Your aim is a balanced profile: no section left far behind.

Practise on real papers

Coaching mocks are useful, but they are written to imitate CAT. The real papers are CAT. Use recent papers, such as CAT 2025 DILR slot 1 or CAT 2024 QA slot 2, as full timed tests, and use the topic-wise practice pages and the test builder to drill specific weaknesses. For a phase-by-phase plan that puts all of this together, read the CAT preparation strategy guide.