Want the number for your own score? The CAT percentile predictor takes a sectional score and returns a band from 45 logged mock results, with every observation behind it published so you can check the working. This page is the analysis behind that tool - why the band exists at all, and why a single-number conversion table cannot be right.
CAT Score vs Percentile: The Short Answer
On the most recent completed cycle (CAT 2025), a total scaled score of about 85 crossed the 99th percentile, 62 crossed the 95th, and 51.5 crossed the 90th. Sectionally, the 99th percentile needed about 44 in VARC, 30 in DILR and 27 in QA. The full table is below.
Two things to hold onto before you use those numbers. They are a benchmark, not a rule: CAT scales scores per slot and the mapping moves every year with paper difficulty and who turned up. And they describe the real CAT - your mock percentiles are a different, noisier distribution, which is what the rest of this piece is about, using 45 of my own scored attempts.
CAT 2025 Score-to-Percentile Table
Figures as reported for the CAT 2025 cycle. CAT's conducting body does not publish a score-to-percentile mapping, so every table like this one - here or anywhere else - is compiled from reported scorecards rather than issued officially. Treat it as the closest available reference point for CAT 2026, not as a cutoff.
| Percentile | Overall | VARC | DILR | QA |
|---|---|---|---|---|
| 99.9 | - | 53 | 38 | 37 |
| 99 | 84.8 | 44 | 29.8 | 27.3 |
| 95 | 62.3 | 32.5 | 21.5 | 18.5 |
| 90 | 51.5 | 26 | 16.7 | 15 |
Source: CAT 2025 scaled-score-to-percentile figures as published by Cracku. The 99.99 percentile that year sat at 132.79 overall. I have not independently verified these against a second published source, and I would rather say so than imply a precision the data does not have.
The Confusion Almost Everyone Has
"I scored 140 in a mock - is that a 99 percentile?" The question has no fixed answer, and that surprises most first-time aspirants. Your score is the raw number from your correct and incorrect answers. Your percentile is relative - the share of test-takers you beat. The two connect only through the difficulty of that specific paper and how everyone else on it performed.
Most articles on this topic stop at that explanation. Below is what it looks like in actual data: 45 score-to-percentile pairs, section by section, from 15 mocks taken by one person between May and August 2026 (the full log is in this write-up of 20 mocks). These are IMS and TIME mock percentiles, not real CAT percentiles - which matters, and I come back to it at the end. But the pattern they show is the point, and it is much stronger than the explanation alone suggests.
The Same Score, Three Different Percentiles
Start with DILR, where the effect is starkest. Every DILR score from those 15 mocks, paired with the percentile it earned:
| DILR score | Percentile | Mock |
|---|---|---|
| 8 | 52 | PreSimCAT 03 |
| 11 | 68 | SIMCAT 3 |
| 13 | 58 | SIMCAT 6 |
| 14 | 59 | SIMCAT 105 |
| 15 | 65 | AIMCAT SA2701 |
| 15 | 76 | SIMCAT 7 |
| 15 | 90 | SIMCAT 103 |
| 16 | 79 | SIMCAT 5 |
| 17 | 87 | SIMCAT 104 |
| 19 | 82 | SIMCAT 101 |
| 20 | 85 | SIMCAT 1 |
| 23 | 75 | SIMCAT 102 |
| 25 | 82 | PreSimCAT 02 |
| 30 | 82 | SIMCAT 2 |
| 30 | 93 | SIMCAT 4 |
Three things in that table are worth sitting with:
- A score of 15 earned the 65th, 76th and 90th percentile on three different papers. Identical performance, a 25-point percentile spread.
- 17 marks beat 23 marks. The 17 landed at the 87th percentile; the 23 at the 75th. Six extra marks cost twelve percentile points, because the paper they were on was easier for everyone.
- A score of 30 got the 82nd percentile once and the 93rd another time. Same score, same section, eleven points apart.
If you had been chasing a percentile target in DILR, this data would have had you celebrating and despairing at essentially the same level of ability.
It Runs the Other Way Too: VARC
The clearest inversion in the whole dataset is in VARC:
| VARC score | Percentile | Mock |
|---|---|---|
| 3 | 12 | PreSimCAT 03 |
| 5 | 15 | SIMCAT 104 |
| 9 | 37 | SIMCAT 7 |
| 10 | 28 | SIMCAT 2 |
| 12 | 42 | SIMCAT 3 |
| 15 | 61 | SIMCAT 103 |
| 17 | 50 | SIMCAT 102 |
| 18 | 55 | SIMCAT 101 |
| 20 | 71 | SIMCAT 5 |
| 21 | 57 | PreSimCAT 02 |
| 25 | 86 | AIMCAT SA2701 |
| 27 | 82 | SIMCAT 1 |
| 33 | 89 | SIMCAT 6 |
| 36 | 92 | SIMCAT 4 |
| 39 | 78 | SIMCAT 105 |
The best VARC score in the set - 39 - earned the worst percentile of the top three. A 39 came 78th; a 33 came 89th; a 36 came 92nd. Six marks better than the 33, and eleven percentile points worse.
Also note 15 → 61st beating 17 → 50th and 18 → 55th, and 20 → 71st beating 21 → 57th. Below the top of the range, the ordering is close to scrambled.
QUANT Shows a Different Failure: the Ceiling
| QUANT score | Percentile | Mock |
|---|---|---|
| 13 | 55 | SIMCAT 5 |
| 19 | 75 | SIMCAT 6 |
| 21 | 80 | SIMCAT 2 |
| 21 | 83 | SIMCAT 3 |
| 23 | 87 | SIMCAT 102 |
| 24 | 91 | SIMCAT 103 |
| 25 | 87 | SIMCAT 104 |
| 25 | 91 | SIMCAT 4 |
| 27 | 83 | SIMCAT 1 |
| 27 | 90 | SIMCAT 105 |
| 28 | 85 | AIMCAT SA2701 |
| 28 | 85 | SIMCAT 101 |
| 30 | 74 | PreSimCAT 03 |
| 36 | 97 | SIMCAT 7 |
| 50 | 97 | PreSimCAT 02 |
Two observations here that neither of the other sections shows:
36 and 50 both earned the 97th percentile. Fourteen extra marks bought nothing at all. Once you are near the top of the distribution, percentile saturates - the marginal mark stops being worth anything, because there is almost nobody left above you.
And 30 earned the 74th percentile, worse than a 24, a 23 and both 21s. That mock was PreSimCAT 03, where the overall was 41 at the 49th percentile - an easy Quant paper on which 30 was unremarkable.
At Overall Level, the Inversions Survive
You might expect these section-level swings to cancel out across three sections. They partly do, but not enough:
- 54 marks → 90th percentile (SIMCAT 103) versus 60 marks → 83rd (SIMCAT 7). Six more marks, seven fewer percentile points.
- 96 marks → 90th percentile (PreSimCAT 02) versus 91 marks → 96th (SIMCAT 4). The highest score in the entire set earned the sixth-best percentile.
- 44 → 70th beat 47 → 67th.
- 65 → 80th and 65 → 82nd - the same score twice, two points apart, which is the closest thing to consistency in the data.
What This Actually Means For Your Prep
Stop treating a single mock percentile as a measurement of you. It measures you against the specific cohort that took that specific paper. Across 15 mocks, the same DILR score spanned 25 percentile points. If you are adjusting your study plan because your percentile fell from 85 to 76, you may be responding to a difference in who else showed up.
Track raw score per section instead. It is noisier in absolute terms but it is yours - it does not move because a different cohort took the paper. The score trend is what tells you whether you are improving.
Read percentile only as a rolling average, and only within one provider. Five mocks from the same platform give you a usable band. One mock gives you nothing, and comparing an IMS percentile against a TIME percentile compares two different candidate pools.
Near the top, chase accuracy rather than marks. The QUANT table shows 36 and 50 tying at the 97th percentile. Above roughly the 95th, extra marks in your strong section are nearly worthless while marks in your weak section are worth a great deal - which is also why sectional cutoffs matter so much (see CAT cutoffs at the top IIMs).
The Caveats, Stated Plainly
These are coaching mock percentiles, and they are noisier than real CAT percentiles for two reasons worth understanding.
First, the cohorts are small and self-selected. A few thousand people take a given SIMCAT, and which few thousand varies by week. Real CAT has upwards of 300,000 candidates, so its distribution is far more stable - you should expect real CAT percentiles to be better behaved than what you see above.
Second, real CAT normalises across slots and mocks generally do not. CAT runs in multiple slots with non-identical papers, and the IIMs convert raw scores into scaled scores to account for slot difficulty, so a 95th percentile in a harder slot represents the same relative standing as a 95th in an easier one. The formula is not published in full, which is one more reason to treat any percentile as an estimate until the official result.
Neither caveat changes the practical conclusion. If anything they sharpen it: the noise in mock percentiles is exactly why your mock percentile is the wrong number to optimise, and your section-wise raw score and accuracy are the right ones.
Tracking the Numbers You Actually Control
None of the patterns above were visible while taking these mocks one at a time. They only appeared once 15 mocks sat in one place with their section scores and percentiles side by side - which is the entire argument for logging mocks at section level rather than recording an overall figure and moving on. Our mock analysis framework covers what to do with each mock; this is what the accumulated set tells you.
Karma Yogi stores each mock as its sections, so score and percentile are separate trend lines per section, and a 25-point percentile swing on a flat score is visible rather than confusing. See how it works for CAT, or start tracking free.
End of essay
- Anish Guruvelli