CAT Percentile Calculation: The Formula, and What It Cannot Tell You
Your CAT percentile is calculated as P = ((N − r) ÷ N) × 100, where N is the total number of candidates who appeared across all slots and r is your rank on the scaled score — not your raw score. The IIMs publish that formula. They do not publish the statistical model that turns a raw score into a scaled one.
That second sentence is the whole story, and most pages on this topic skip it. It means you cannot compute your own CAT percentile, and every percentile predictor online — including any we could build — is an estimate fitted to past years, not a calculation. Below is exactly what is knowable, what is not, and what to do with the difference.
Raw score → scaled score → percentile
Step one converts your raw marks into a scaled score to make different slots comparable. Step two ranks everyone by scaled score and converts rank to percentile. Step two is public arithmetic. Step one is not, and that is where the uncertainty in every prediction lives.
Quick Answer (30-Second Read)
The formula
P = ((N − r) ÷ N) × 100
Official, published by the IIMs
The catch
r is your rank on the scaled score
Scaling model is not published
- Percentile is relative, not absolute. It measures where you stand among everyone who appeared, not how many marks you scored.
- Raw score is never used directly. Marks are scaled first so that slots of different difficulty are comparable.
- A hard slot is not a disadvantage. Scaling is what removes the disadvantage — that is the entire reason it exists.
- You cannot calculate your percentile yourself. The scaling model is not public, so predictors are fitted estimates.
- Sectional percentiles are calculated the same way, separately. Most schools screen on those too.
Formula and scaling policy per the official CAT information bulletin — verify your year's version at iimcat.ac.in.
CAT Mock Test — VARC, DILR & QA on the real interface
Sections close on time exactly like the real paper. Your section-wise report opens the moment you submit — including how many TITA questions you left blank.
The percentile formula
This part the IIMs publish, and it is ordinary arithmetic:
P = ((N − r) ÷ N) × 100
N — the total number of candidates who appeared in CAT across all slots.
r — your rank among those candidates, ordered by scaled score.
P — your percentile, rounded to two decimal places.
Read what that formula does not contain: your marks. Nowhere in the calculation does a raw score appear. Percentile is built from rank, and rank is built from scaled score. Marks matter only insofar as they produce a rank.
This is why the question "how many marks for 99 percentile?" has no fixed answer, and why anyone who gives you one without a range is guessing. The mark that produces a given percentile changes every year with the difficulty of the papers and the size and strength of the candidate pool.
Ties are handled by rank, not by mark
Candidates with identical scaled scores receive the same rank, which means they receive the same percentile. A cluster of tied candidates therefore occupies one rank position and the next distinct scaled score jumps past all of them — one reason percentiles bunch up densely in the middle of the distribution and spread out at the top.
A worked calculation
The figures below are illustrative — chosen to show the mechanics clearly, not taken from any particular year. Real values of N vary by cycle.
Setup. Suppose 3,00,000 candidates appeared, and your scaled score places you at rank 3,000.
Step 1. N = 3,00,000 · r = 3,000
Step 2. N − r = 3,00,000 − 3,000 = 2,97,000
Step 3. 2,97,000 ÷ 3,00,000 = 0.99
Step 4. 0.99 × 100 = 99.00
Percentile: 99.00
Two things worth noticing in that arithmetic, because both surprise people.
| Rank | Percentile (N = 3,00,000) | Candidates ahead of you | What moving up costs |
|---|---|---|---|
| 30,000 | 90.00 | 29,999 | — |
| 15,000 | 95.00 | 14,999 | Overtake 15,000 people |
| 3,000 | 99.00 | 2,999 | Overtake another 12,000 |
| 1,500 | 99.50 | 1,499 | Overtake another 1,500 |
| 300 | 99.90 | 299 | Overtake another 1,200 |
First: the gap between 99.00 and 99.90 is only 2,700 people, while the gap between 90.00 and 95.00 is 15,000. Percentiles at the top are compressed — a handful of additional marks moves you a long way, and a handful of careless errors costs you just as much.
Second: because N changes each year, the same rank produces a different percentile in different cycles. A rank of 3,000 is 99.00 when 3,00,000 appear and 98.75 when 2,40,000 appear. Nothing about your performance changed.
Why scaling exists at all
CAT runs in multiple slots on the same day, and each slot gets a different paper. Those papers cannot be identical in difficulty — no amount of care in question setting achieves that. So comparing raw marks across slots would be unfair to whoever drew the harder paper.
Scaling solves this by converting raw scores in each slot onto a common measurement scale, so that a given scaled score represents the same level of performance regardless of which paper produced it. Only then are candidates ranked together.
| Raw score | Scaled score | Used for percentile? | |
|---|---|---|---|
| What it is | Marks you actually earned | Your performance placed on a common scale | — |
| Comparable across slots | No | Yes | — |
| Shown on your scorecard | Yes | Yes | — |
| Feeds the percentile | No | Yes | Scaled only |
The practical consequence is one you should act on: stop reading slot-difficulty discussions after your exam. If your paper was harder, everyone in your slot found it harder, and scaling is precisely the mechanism that accounts for it. The anxiety those threads generate is about a problem that has already been solved.
Estimates are only as good as the mock behind them
A percentile estimate built on a full-length timed paper with the real 40-minute lock is worth something. One built on an untimed practice set is not.
- Percentile estimate
- Sectional breakdown
- Cohort comparison
What the IIMs do not publish
Here is the part that most pages on this topic either skip or actively obscure, because it is inconvenient for anyone selling a predictor.
The statistical model used to convert raw scores into scaled scores is not made public. The IIMs state that a normalisation and scaling procedure is applied, and they publish the percentile formula that follows it. They do not publish the parameters, the equating method's specifics, or the slot-level data the model runs on.
Which means, stated plainly:
| Question | Can you answer it yourself? | Why |
|---|---|---|
| Given a rank and N, what is my percentile? | Yes | Public formula, simple arithmetic |
| Given my raw score, what is my scaled score? | No | Model and parameters are not published |
| Given my raw score, what is my rank? | No | Requires everyone else's scaled scores |
| Given my raw score, what is my percentile? | No | Depends on both of the above |
So when a site offers to calculate your CAT percentile from your raw score, what it is actually doing is fitting a curve to previously reported score-percentile pairs and reading your number off that curve. That can be genuinely useful as a rough guide. It is not a calculation, and it should not be described as one.
Why two percentile predictors give you different answers
If you have put the same raw score into three predictors and received three different percentiles, nothing is broken. They are estimates built on different assumptions.
- Different reference years. A predictor fitted on a cycle where papers ran harder will map your score to a higher percentile.
- Different assumptions about N. Candidate numbers move year to year, and the same rank maps to different percentiles at different N.
- Different self-reported data. Predictors trained on user-submitted scores inherit whatever bias sits in who chooses to submit — and strong scorers report more readily than weak ones.
- Different handling of sectionals. Some estimate an overall percentile only; others model each section, which changes the result for lopsided profiles.
How to use a predictor sensibly
Treat the output as a band, not a number. If three predictors return 96.2, 97.8 and 98.4, your honest read is "high nineties, probably not 99+". That band is genuinely useful for shortlisting which schools to apply to. Reading it as a decimal figure and planning around it is where people go wrong.
Sectional percentiles work the same way — and matter more than people expect
Each section is scaled and ranked separately, then the same percentile formula is applied. You therefore receive four percentiles: one for VARC, one for DILR, one for QA, and one overall.
The overall percentile is not an average of the three. It is calculated from your overall scaled score, which is why a candidate can hold a higher overall percentile than any of their individual sectional percentiles, or lower than the highest of them.
Most business schools that use CAT apply a minimum sectional percentile alongside the overall requirement. The practical effect is that a lopsided profile is punished harder than the overall number suggests:
| Profile | VARC | DILR | QA | Typical outcome |
|---|---|---|---|---|
| Balanced | 96 | 95 | 97 | Clears sectional screens comfortably |
| Lopsided, higher overall | 78 | 99 | 99 | Often filtered out on the VARC screen despite a stronger total |
| Weak in one, strong elsewhere | 99 | 82 | 96 | Depends entirely on each school's DILR floor |
This is the single most actionable consequence of how CAT percentiles are built: past a point, raising your weakest section is worth more than raising your strongest. Ten extra marks in a section where you sit at the 78th percentile move you far further than ten marks where you already sit at the 99th.
Key Takeaways
- P = ((N − r) ÷ N) × 100 — the formula is public and simple, and it uses rank, never marks.
- Rank comes from your scaled score, and the model that produces scaled scores is not published.
- Nobody can calculate your CAT percentile, including us. Every predictor is a fitted estimate.
- Percentiles compress at the top — the jump from 99.00 to 99.90 is only about 2,700 ranks.
- A hard slot is not a disadvantage. Scaling exists precisely to remove it.
- Fix your weakest section first. Sectional screens punish lopsided profiles more than the overall number suggests.
People also search for
How is the CAT percentile calculated?
The IIMs use P = ((N − r) ÷ N) × 100, where N is the total number of candidates who appeared across all slots and r is your rank ordered by scaled score. The result is rounded to two decimal places. Note that raw marks appear nowhere in this formula — percentile is built entirely from rank, and rank is built from the scaled score rather than the marks you actually earned. This is why the same raw score produces different percentiles in different years.
Can I calculate my CAT percentile myself?
Not from your raw score, no. You can apply the percentile formula perfectly well if you already know your rank and the total number of candidates, but you cannot get from marks to rank on your own. That step requires the scaled score, and the statistical model that converts raw scores into scaled ones is not published by the IIMs. Any tool claiming to calculate your percentile from marks is fitting a curve to previously reported data, which is an estimate rather than a calculation.
What is the difference between raw score and scaled score in CAT?
Your raw score is the marks you actually earned on your paper. Your scaled score is that performance placed onto a common measurement scale so it can be compared fairly against candidates who sat a different slot with a different paper. Both appear on your scorecard, but only the scaled score is used to rank candidates and therefore to produce percentiles. Two candidates with identical raw scores in different slots can end up with different scaled scores if their papers differed in difficulty.
How many marks are needed for 99 percentile in CAT?
There is no fixed answer, and treat any source that gives one without a range with caution. The mark required for a given percentile depends on how difficult that year's papers were and on the size and strength of the candidate pool, both of which move every cycle. Because percentile is calculated from rank rather than marks, the same score can be 98.6 one year and 99.2 the next. Aim at a rank band rather than a mark target, and use mock percentile estimates as a directional guide.
Does a difficult CAT slot lower my percentile?
No, and this is exactly what scaling exists to prevent. If your slot's paper was harder, every candidate in that slot found it harder, and the scaling process adjusts raw scores so that performances across slots become comparable before anyone is ranked. Your percentile reflects your standing among all candidates who appeared, not your raw mark count. Reading post-exam threads about which slot was toughest generates anxiety about a problem the process has already handled.
Why do different CAT percentile predictors give different results?
Because each is an estimate built on different assumptions rather than a calculation. Predictors differ in which past year they fit their curve to, what they assume the total candidate count will be, whether their training data comes from self-reported scores, and whether they model sections separately or only the overall score. If three predictors return 96.2, 97.8 and 98.4 for the same input, the honest reading is a band — high nineties, probably not 99 plus — not any one of those decimals.
Is the overall CAT percentile an average of the sectional percentiles?
No. Your overall percentile is calculated from your overall scaled score using the same formula applied to the full candidate list, not by averaging your three sectional percentiles. This is why your overall percentile can sit above all three of your sectional percentiles, or between them, in ways an average would not produce. Each section is scaled and ranked independently, so you receive four separate percentile figures on your scorecard.
Do sectional percentiles matter for IIM calls?
Yes, and often more than candidates expect. CAT itself does not set a sectional cut-off, but most business schools that use the score apply a minimum sectional percentile alongside their overall requirement. A candidate at 99 in DILR and QA but 78 in VARC is frequently filtered out at the sectional screen despite holding a strong overall percentile. Past a certain point, raising your weakest section is worth considerably more than adding marks to your strongest.
How are ties handled in CAT percentile calculation?
Candidates with identical scaled scores are assigned the same rank, and therefore receive the same percentile. The next distinct scaled score then takes the rank position after the entire tied group. This is one reason percentiles bunch densely in the middle of the distribution, where many candidates cluster around similar scores, and spread out towards the top, where scores are more thinly distributed and a single mark separates more rank positions.
How much does one mark change my CAT percentile?
It depends entirely on where you sit in the distribution. Near the middle, where candidates are densely clustered, a single mark can move you past a large number of people and shift your percentile noticeably. Near the top, ranks are more thinly spread — the gap between 99.00 and 99.90 is only around 2,700 ranks in a pool of three lakh — so each mark matters more per rank but the percentile movement per mark is smaller in absolute terms. This is why careless errors are so expensive at the top end.
Like what you read? Share with a friend
Senior Content Writer
Co-Founder at PrepGrind, working on education content, product development, and student-focused learning resources. I also write articles and guides for MBA/MBS entrance exams, covering preparation strategies, exam updates, and useful resources for students.
