How CAT Turns Raw Marks Into a Scaled Score and a Percentile

CAT is sat in three sessions with three different papers, so your raw marks are never compared directly with anyone else’s. The published calculation first equates each session’s scores to a common scale, caps each section at 100, adds the three sections, and turns the result into a percentile using your rank among everyone who appeared. The IIMs also state that scores are “further normalised across different sections” by percentile equivalence, a step they do not describe in enough detail to reproduce (Step 3 below). The method is published — the CAT 2024 and CAT 2025 documents are identical in substance — but almost every “CAT percentile calculator” online skips it. This guide walks through the official formula, lets you run it, and checks it against the numbers the IIMs actually announced.

CAT 2025 result release (24.12.2025) and Scaled Score Calculation.

The raw score is calculated per section: VARC, DILR and QA each get their own. Two objections were accepted in CAT 2023 — one VARC answer key was corrected and one DILR question was given two correct options — which is a reminder that the raw score is only final after the answer-key review.

For every section, the CAT Centre computes two numbers for each session and the same two numbers for all sessions combined. The first is G = mean + standard deviation of the raw scores, a marker for a good-but-typical performance. The second is the mean raw score of the top 0.1%, a marker for the very top. The formula then maps your session’s scale onto the combined scale so that those two markers line up.

Read the formula as “how far above your session’s typical mark were you, measured in units of your session’s spread to the top?” — then re-expressed on the combined scale. The official note says the aim is to “adjust for location and scale differences of score distributions across different forms”. Whether a given raw score moves up or down depends on both anchors of your session (Gi and the top-0.1% mean) against the combined ones — a simple “hard session” or “easy session” label does not decide it, and the formula does. A candidate exactly at their session’s Gi gets exactly the combined G.

If the formula gives more than 100 in a section, the scaled score is rounded down to 100. The overall scaled score is simply the sum of the three sectional scaled scores. The CAT 2025 Scoring and Equating note adds one sentence the calculation document does not expand on: after normalising across forms, “the scores shall be further normalised across different sections”, using “percentile equivalence”. The published step-by-step calculation describes only the session equating and the sum, so we cannot tell you how that cross-section step is applied — both statements are official, and we print both rather than guess.

Because percentile comes from rank, it depends on how many people appeared. The rounding rule is also why “100 percentile” is not one person: at 2.58 lakh candidates, every rank from 1 to 12 rounds to 100.

Every result release lists how many candidates scored 100, 99.99 and 99.98 percentile. If the formula above is really what the IIMs use, those counts should be exactly the ranks that round to each value, given the number who appeared. We ran that check. Using the rounded appeared figures as approximations for N, all nine counts match. Because the exact N and the tie pattern are not published, this is supporting evidence, not an exact verification.

Turn the formula round and a percentile gives an approximate rank: r ≈ N × (1 − P/100); the number strictly ahead of you is r − 1. Because the displayed percentile is rounded to two decimals, it really stands for a range of ranks (at 2.58 lakh, a displayed 99.99 is any rank from 13 to 38) — the calculator below shows that range. The table gives central estimates for the last two fields. It is arithmetic, not an official rank list, and the 2026 field is unknown until the result — but it shows why the same percentile means a different number of people each year.

The IIMs publish the method, not the inputs. There is no official table that says “this many marks gives this percentile”: the session means, standard deviations and top-0.1% averages stay inside the CAT Centre, and the percentile depends on the whole field. Third-party marks-vs-percentile tables are estimates assembled from sources such as candidates’ response sheets or scorecards, score-calculator submissions and previous-year trends; the method varies by publisher. Such tables can be useful for a rough target, but they are estimates on scaled scores from a previous year, and a table “out of 204” also assumes a 68-question paper, which the IIMs have not announced.

What you can use with confidence: your raw score is +3/−1 on MCQs and +3/0 on typed answers; session differences are adjusted by the equating; each section is capped at 100 scaled; and your percentile is your rank among everyone who appeared, rounded to two decimals. For an estimate of rank from a percentile, the CAT rank predictor in this series does the same arithmetic in more detail.