CAT Normalisation Explained: Why Your Scaled Score Isn't Your Raw Score
Three sessions, three different papers, one merit list. Normalisation is the statistical step that makes that possible, and it sits between the marks you earned and the percentile you are ranked on.
It is also the source of more misplaced anxiety than any other part of the process. The short version: normalisation is designed so that your slot does not matter, and the time spent worrying about which slot was harder is time that would earn more anywhere else.
Same percentile, same scaled score, whichever slot you sat
The stated aim of equipercentile normalisation is that a candidate at the 90th percentile of the morning session and one at the 90th percentile of the evening session land on the same scaled score — even though their raw marks may differ substantially.
Quick Answer (30-Second Read)
What it fixes
Slot difficulty
Not your performance
Order of operations
Raw → scaled → %ile
Percentile comes last
- Equipercentile normalisation maps each slot's score distribution onto one common scale.
- Two steps: adjust for differences between slots, then normalise each section.
- Your scaled score can be above or below your raw score — both are normal.
- Percentiles are computed on scaled scores, which is why your own estimate never matches exactly.
- You cannot choose or game your slot. The whole mechanism exists so you would not benefit if you could.
The normalisation methodology is described in the official CAT bulletin and score documentation at iimcat.ac.in.
CAT Mock Test — the part of the score you control
Normalisation is not something you can prepare for. Attempts, accuracy and section timing are. Start there.
Why it exists at all
A single national exam taken by hundreds of thousands of candidates cannot be delivered in one sitting. Splitting it into sessions is a logistical necessity — and the moment you split it, you need different question papers, because a paper reused in the afternoon is a paper that has already circulated.
Different papers are never equally difficult. Without a correction, a candidate who happened to sit the harder session would be penalised for something entirely outside their control. Normalisation is the correction.
| Without normalisation | With normalisation |
|---|---|
| Raw marks compared directly across slots | Marks converted to a common scale first |
| The easier paper's candidates rank higher as a group | Rank within your own slot is what carries forward |
| Slot allotment becomes a lottery worth gaming | Slot allotment becomes irrelevant to your outcome |
| A 70 in a brutal paper loses to a 78 in an easy one | Both are placed by where they stood among their peers |
How it works, in plain language
The method used is equipercentile normalisation. The name sounds forbidding and the idea behind it is not.
Take the full distribution of raw scores from each session. Rather than comparing marks to marks, compare positions to positions. The candidate at the 85th percentile of the morning session and the candidate at the 85th percentile of the afternoon session are treated as equivalent performances, and both are assigned the same scaled score — regardless of whether one earned 74 raw marks and the other 81.
The two steps, in order
First, scores are adjusted for the differences between the distributions of the three sessions, placing everyone on a shared scale. Second, the same treatment is applied section by section, so VARC, DILR and QA are each normalised in their own right rather than only in aggregate. That second step is why your scorecard reports a scaled score for every section separately.
The published formula expresses this as a linear transformation — the scaled score is the mean of the target scale, plus your distance from your own slot's mean, rescaled by the ratio of the two standard deviations. Which is a compact way of saying: your position within your own group is what carries forward, not your marks.
A worked example
Illustrative numbers, chosen to make the mechanism visible. They are not figures from any actual cycle.
Two candidates, two sessions.
Candidate A sits a hard morning paper and scores 68 raw. The morning mean is 42. Candidate B sits an easier afternoon paper and scores 76 raw, against an afternoon mean of 53.
Compare marks and B wins by eight. Compare positions and A is 26 marks above their slot's mean while B is 23 above theirs — a closer contest than the raw gap implies, and one that can invert once the spread of each distribution is taken into account.
Normalisation resolves it by rank, not by marks. If A stood at the 96th percentile of the morning session and B at the 94th of the afternoon, A finishes ahead despite the lower raw score.
The consequence: a hard paper does not hurt you, because everyone in that room faced it. What matters is how you did relative to them.
This is also why "my slot was brutal" is rarely the explanation for a disappointing percentile. If the slot was brutal for you, it was brutal for everyone sitting with you, and the normalisation absorbs precisely that.
Work on the part you can actually move
Slot difficulty is handled for you. Attempts, accuracy and time distribution are not — and a full paper measures all three.
- Sectional percentiles
- Accuracy vs attempts
- Time per question
Raw, scaled, percentile — three different numbers
| Number | Where it comes from | Do you see it? |
|---|---|---|
| Raw score | (Correct × 3) − (Wrong MCQs × 1), no penalty on TITA | Only if you compute it from the response sheet |
| Scaled score | Raw score after normalisation across slots and sections | Yes — this is what the scorecard reports |
| Percentile | Your rank position, computed on the scaled score | Yes, overall and section-wise |
The sequence matters. Percentile is computed after scaling, not before — which is why a candidate who calculates their raw score from the answer key and looks it up in a historical table gets a useful band and never an exact figure. The scaling step sits in between, and it has not been applied yet at the moment you do that arithmetic.
A scaled score higher than your raw score means your session's distribution sat below the common scale; lower means the reverse. Neither is a reward or a punishment. Both are the same adjustment working in different directions.
Four things people believe about slots
| Belief | What is actually the case |
|---|---|
| "The third slot is easier" | Difficulty varies unpredictably and is corrected for regardless. Even if true in a given year, you could not have acted on it |
| "Toppers cluster in one slot" | Strong candidates appear across all sessions; the allotment is not selective |
| "Normalisation reduces my marks" | It moves scores in both directions depending on your session's distribution |
| "I can request a preferred slot" | No. It is allotted and printed on the admit card |
There is a practical instruction buried in all four rows, and it is the only one this topic yields: since you cannot know your session until early November, rotate your mock timings across morning, afternoon and evening from October onward. That is the sum total of what a candidate can do about slots — and it is worth more than any amount of speculation about which one is easier.
Key Takeaways
- Normalisation exists because three sessions need three papers, and three papers are never equally hard.
- The equipercentile principle compares positions, not marks — same percentile in any slot, same scaled score.
- Two steps: across slots first, then section by section — which is why every section gets its own scaled score.
- Percentile is computed after scaling, so a raw-score estimate gives you a band and never a decimal.
- A hard paper does not hurt you — everyone in that session faced it, and that is exactly what gets absorbed.
- The only actionable response is rotating mock timings from October, since the slot arrives with the admit card.
People also search for
What is CAT normalisation?
A statistical process that converts raw marks from three differently-worded sessions onto a single common scale, so that candidates can be ranked against one another fairly. It uses an equipercentile approach: rather than comparing marks to marks, it compares positions to positions, so that a candidate at the 90th percentile of one session and a candidate at the 90th percentile of another receive the same scaled score even when their raw marks differ substantially.
Why does CAT need normalisation?
Because the exam cannot be delivered to hundreds of thousands of candidates in one sitting, and the moment it is split into sessions each session needs a different paper — a paper reused later in the day is one that has already circulated. Different papers are never equally difficult, so without a correction a candidate who happened to sit the harder session would be penalised for something entirely outside their control.
How is the CAT scaled score calculated?
Through a two-step process. First, scores are adjusted for differences between the distributions of the three sessions, placing all candidates on a shared scale. Second, the same treatment is applied section by section, which is why the scorecard reports a separate scaled score for VARC, DILR and QA. The published formula expresses this as a linear transformation of your distance from your own session's mean, rescaled by the ratio of standard deviations.
Can normalisation reduce my CAT score?
Your scaled score can come out below your raw score, yes — and it can equally come out above it. Which direction it moves depends on how your session's score distribution sat relative to the common scale, not on anything about your performance. Neither outcome is a reward or a punishment; both are the same adjustment operating in different directions, and both leave your position relative to the candidates who sat with you unchanged.
Is one CAT slot easier than the others?
Difficulty does vary between sessions, unpredictably, and normalisation exists specifically to correct for it. Even if one session turned out easier in a given year, no candidate could have acted on that, because slots are allotted rather than chosen and appear only on the admit card in early November. The one useful response is to rotate mock timings across morning, afternoon and evening from October so that any allotted session feels ordinary.
Why doesn't my calculated score match my scorecard?
Because the two are different quantities separated by the normalisation step. What you compute from the response sheet and answer key is a raw score — three marks per correct answer, minus one per wrong MCQ, nothing deducted for wrong non-MCQ answers. What the scorecard reports is the scaled score produced after normalisation, and your percentile is computed on that. A raw-score estimate is genuinely useful for placing yourself in a band, and never accurate to a decimal.
Does normalisation apply to each section separately?
Yes. The second step of the process normalises VARC, DILR and QA in their own right rather than only in aggregate, which is why your scorecard carries a scaled score and a percentile for each section as well as overall. This matters practically, because institutes screen sectional percentiles against sectional minimums — so the section-level normalisation feeds directly into whether your application clears a filter.
Does a harder CAT paper mean a lower percentile?
No, and this is the most useful thing to understand about the whole mechanism. If the paper was hard for you, it was hard for everyone sitting the same session, and normalisation absorbs exactly that. What a harder paper does change is the marks needed for a given percentile across the cycle as a whole — which is why the score required for 99 percentile moves by several marks between years without anything unusual having happened.
Can I see how my score was normalised?
Not in individual detail. The methodology is described in the official documentation, but the session-wise distributions and the parameters used in a given cycle are not published candidate by candidate, and there is no mechanism to request a breakdown of your own adjustment. What you receive is the scaled score and the percentile it produced. Any third-party tool claiming to reverse-engineer your normalisation is estimating, not reporting.
Should normalisation change how I prepare?
Only in one narrow respect: rotate your mock timings across all three session windows from October, since you will not know your allotted slot until the admit card releases. Beyond that, no. Normalisation is a correction applied after the exam to something you cannot influence, and attention spent on it is attention taken from attempts, accuracy and section timing — the three things that actually determine where you land in your own session's distribution.
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.
