CAT Slot Normalization: How Raw Scores Become Percentiles

CAT is conducted across three slots on exam day - morning, afternoon and evening - each with a different question set, and no two of those papers are ever exactly equal in difficulty. If IIMs simply ranked every candidate by raw score regardless of slot, a candidate who happened to sit an easier paper would have a real, unearned advantage over one who sat a harder one. Normalization is the statistical step that removes that advantage: percentile is computed relative to the specific slot a candidate sat, not against the whole candidate pool on a common raw-score scale.

What follows is why that step exists, how equating works as a general statistical method in large-scale testing, and a plain statement of the thing most explanations of this topic skip past: the exact formula IIMs use has never been officially published, and every specific formula circulating online - including ones stated with total confidence - is a third party's reconstruction, not a confirmed methodology.

Why slot difficulty is a real problem, not a theoretical one

CAT's slots do not share a question paper. Each slot sits a different set of VARC, DILR and QUANT questions, assembled to a broadly similar difficulty target but never identical in practice - one slot's DILR sets might include a harder arrangement puzzle, another slot's QUANT might lean more heavily on a topic that candidate pool happens to be stronger in. Difficulty also depends on who is in the room: a slot with a stronger cohort compresses the score distribution differently from one with a weaker cohort, even on an identical paper.

Both effects mean a raw score of, say, 21 in DILR carries a different amount of information depending on which slot produced it. Comparing raw scores directly across slots would silently reward whoever happened to sit the easier paper or the weaker cohort - which is exactly the problem normalization exists to solve.

What IIMs have actually disclosed

This is the part worth being blunt about. Searching current CAT guidance and a wide range of coaching-industry explainers turns up a consistent pattern: institutes describe the existence and purpose of normalization confidently, and then, without exception in what was found, concede that the precise statistical formula is not officially published by IIMs. One representative summary states plainly that "the exact statistical formula is not disclosed by IIMs, but it's based on statistical normalization" - and that gap between "we know it happens" and "we know exactly how" runs through every source consulted for this piece.

So when you see a specific formula presented online as CAT's normalization method - a particular mean-and-standard-deviation scaling equation, a named equating technique - treat it as a reconstruction or a simplification, not an official disclosure. IIMs communicate the outcome (percentiles, category-wise cutoffs) and the general principle (scores are adjusted for slot-difficulty differences before percentiles are computed), not the underlying computation.

How equating actually works, in principle

What can be stated with confidence is the general family of methods this kind of problem is normally solved with in large-scale testing, because these are standard, published techniques used across many exams worldwide (the SAT and AP exams among them), not something specific to CAT. Test equating is the general term for statistically adjusting scores from different test forms so they can be compared on a common scale. Two common approaches:

• Linear (mean/SD) equating adjusts a raw score so that the two forms end up with a comparable mean and standard deviation - it shifts and rescales the whole distribution by a single factor.
• Equipercentile equating maps scores so that whatever percentile you sat at within your own form's distribution is treated as equivalent to the same percentile on another form. This can produce a non-linear relationship between raw score and scaled score, because it directly matches rank-within-form rather than assuming both distributions have the same shape.

Both are legitimate, well-established statistical tools. Neither is confirmed as CAT's actual method - both are simply the standard menu that a real answer would be drawn from, and different coaching explainers reach for different ones without any of them citing an IIM source for the choice.

A worked illustration - hypothetical, not IIM's real data

To make the mechanism concrete rather than abstract, here is a constructed example using an equipercentile-style approach. The numbers are invented for illustration only; they are not reported IIM figures.

Same raw score, same section, an 11-point percentile gap - purely because the two candidates were compared against different rooms of people on different papers. This is the mechanism, illustrated with invented numbers; the direction of the effect (identical raw scores landing at different percentiles depending on slot) is real and is exactly what this site's own logged mock data shows happening even without official slot normalization - see 45 real score-to-percentile pairs from 15 mocks and 18 logged attempts compared for the actual observations. Those two posts are about the symptom; this one is about the mechanism that makes the symptom structurally unavoidable in a multi-slot exam.

Why mock analysis needs this caveat, specifically

Coaching mocks usually do not attempt slot normalization at all - most run one paper for everyone rather than multiple slots with different question sets, so there is nothing to normalize across in the first place. That is a genuine, separate reason mock percentiles behave unpredictably (covered in the two posts linked above), distinct from the slot-normalization question this post is about. But the underlying lesson compounds: real CAT percentiles are the output of a genuine cross-slot equating process whose exact method is undisclosed, while most mock percentiles are not equated across anything at all. Neither number is safe to read as a precise, portable measurement of your ability - for two different, stackable reasons.

What this means in practice

• A raw-score comparison between two candidates in different slots is not meaningful on its own. Only their percentiles, computed within their own slot, are comparable - and even that comparison assumes IIMs' undisclosed method behaves the way the standard equating techniques above would predict.
• Do not trust a specific formula presented online as CAT's official method. Every source found for this piece that states one is presenting a reconstruction, not a citation to an IIM publication.
• Treat your own raw sectional scores as the number you control, and your percentile as an estimate produced by a process you cannot fully audit - the same practical advice this site's percentile posts already give, for a related but distinct reason.
• Slot allocation is outside your control, so there is no strategic action to take here beyond understanding why two equally-prepared candidates can land at different percentiles through no fault of either one's preparation.

Where this is weak

• IIMs have not published the exact normalization formula, and this post does not claim to reveal it. Everything under "how equating actually works" describes standard, general-purpose psychometric methods used in large-scale testing broadly - not a confirmed description of CAT's specific computation.
• The worked table above is invented for illustration. It is not drawn from released IIM data, and no such per-slot percentile breakdown has been published for any real CAT cycle to my knowledge.
• The "not disclosed" claim rests on secondary sources, not a primary IIM statement saying so. Every coaching-industry explainer consulted converges on the same conclusion, which is meaningful convergence, but none of them is IIM's own technical documentation, because no such public document appears to exist.
• Equipercentile and linear equating are presented as the standard menu, not as a confirmed choice. IIMs could be using either, a hybrid, or something else entirely; nothing here should be read as narrowing that down.
• This reasoning has not been checked against an actual released CAT percentile dataset, because IIMs do not release slot-level raw score distributions publicly.

Sources

• Wikipedia, Test equating, read in full 2026-09-10, for the general definitions of linear and equipercentile equating and their use in large-scale exams such as the SAT and AP tests.
• Coaching-industry explainers on CAT normalization (including Tarkashastra's slot-normalization explainer, read in full) converge on the same statement - that IIMs have not disclosed the exact formula - without citing an official IIM publication for it; treat this as convergent secondary evidence rather than a primary source.
• CAT's three-slot exam-day structure and section timings, as established in this site's pattern and syllabus post.

None of this changes what is actually within a candidate's control: the raw sectional score. Karma Yogi stores each mock's per-section score, attempts and percentile separately by provider, so a percentile swing that turns out to be about the paper rather than about you stays visible as exactly that, rather than triggering a study-plan change it does not deserve.