How to Analyse CAT Mocks: 5 Steps, One 96-40-94 Week

This is how I analyse a CAT mock - five steps, worked through three real papers from my own log. (If you spell it "analyze", same thing; the method does not change.) The short version: the score is the least useful number on the report, and the section breakdown three mocks either side of it is where the decision actually lives.

The mock that should have worried me, and didn't

Three of my full mocks, in order: 96 on 29 July, 40 on 30 July, 94 on 1 August. A 56-mark drop followed by a 54-mark recovery, inside four days. If you only look at the middle number, it looks like a collapse worth rebuilding a study plan around. Looked at as a three-mock stretch, it looks like one bad day sandwiched between two good ones - and the section breakdown backs that reading up:

All three sections dropped on 30 July, not just one. That pattern - a whole-mock dip rather than a single weak section - is itself diagnostic: it points at something that day (fatigue, a bad start, poor sleep) rather than a content gap in any specific area. A section-specific crash would have looked different: two sections holding steady while one section alone fell off a cliff. This one didn't.

One honest limitation: I don't have a per-question time log for the 30 July mock, so I can't say with certainty whether it was a time-management collapse (panicking early, then rushing everything after) or something else entirely. That absence is itself worth naming - without a time log, a mock analysis has a real gap in it, and "I don't know" is a more useful entry than a guessed one.

Step 1: classify every wrong or skipped question by error type

Before anything else, tag every wrong or skipped question as one of three types:

• Conceptual: you didn't know the method. A knowledge gap - fixable by revisiting the topic.
• Careless: you knew the method but slipped - arithmetic error, misread question, mis-transcribed number. A discipline gap, not a content gap.
• Strategic: you attempted something that, in hindsight, should have been skipped given the time cost, or skipped something that was actually quick.

A real example makes the distinction concrete. On 17 May I sat a DILR sectional and scored 18/66 - I attempted 8 questions and got 6 of them right. Six out of eight is 75% accuracy, which rules out a conceptual gap; whatever I attempted, I mostly solved. A later revision pass (logged as R2) flagged that the set actually supported 39 marks, with a specific note pointing at Q10, a Games and Tournaments question. The gap between an actual 18 and a possible 39 isn't a knowledge problem at all - it's a strategic one: too few questions attempted, not too many solved wrong. Classifying it correctly is what tells you the fix is "attempt more of what you can already do," not "study Games and Tournaments harder."

What "logged, but not yet analyzed" actually costs

Across 23 full mocks and 6 sectionals, the ones that changed how I studied were never the ones I just took and moved past - they were the ones I came back to with the classification above applied honestly. It's tempting, especially after a good score like the 96 on 29 July, to skip the analysis step entirely: the mock felt fine, so what is there to find? The 06-14 VARC sectional answers that directly - a score that felt reasonable (19/72, 81st percentile) turned out to have a documented 35 sitting inside it, found only because a second pass happened at all. A good score is not evidence that analysis would find nothing; it's just evidence that whatever was missed didn't cost enough to be visible from the score alone.

Step 2: a section-wise time audit, where you have the data

Most mock platforms log time spent per question. Where that log exists, the questions worth asking are: did you spend disproportionate time on something you got wrong anyway? Did you run out of time with easy, attemptable questions left unanswered? This tells you whether your set-selection or question-ordering habits (see in-exam time management) actually held up under real pressure, or only in theory. Where the log doesn't exist - as with my 30 July mock - say so plainly rather than reconstructing a time story from memory, which is usually wrong in a self-serving direction.

Step 3: topic-level accuracy, not just section-level

"My DILR was weak this mock" is too vague to act on. Compare 22 August (DILR 23, overall 62, 92nd percentile) against 24 August (DILR 20, overall 72, 96th percentile) two days later: DILR actually dipped slightly while VARC (14 → 20) and QUANT (25 → 32) both improved enough to carry the overall score and the percentile up. Read at the overall level, that pair looks like unambiguous improvement. Read at the section level, it's improvement in two sections and a small step back in a third - which is the more useful thing to know, because it tells you DILR specifically didn't benefit from whatever changed between those two mocks.

Step 4: the revision round, and why the second pass found what the first missed

A single analysis right after the mock isn't enough - the 39-versus-18 DILR gap wasn't visible on the day of the test. It showed up on a second pass, logged as R2, once the emotional pull of the original score had faded enough to look at the set more honestly. Karma Yogi's revision round system tags each review R1 through R7 against the mock or sectional it belongs to, on a spaced schedule - R1 around 3 days out, R2 around a week, and further passes out to R7. My overall study plan covers where this sits in the wider prep timeline.

Step 5: track the trend, not the single data point

This is the 96-40-94 stretch again, stated as a rule: no single mock score means much on its own. Difficulty, energy, and plain randomness all add noise to one attempt. What matters is the run of scores around it - a single bad mock between two good ones is usually noise; a specific topic accuracy that keeps dropping across three or four mocks running is signal. Confusing the two is the most common mistake in mock analysis, and it runs in both directions: a 96 the day before a 40 doesn't mean the 40 was a fluke to ignore, and a 40 doesn't mean the 96 was a fluke either. The stretch as a whole is the actual data point.

The dimension most analysis frameworks skip entirely: how the session felt

Every session in my log carries a mood rating alongside its duration - my average across 340 sessions is 3.60 out of 5. That's not a mock-specific number, but it's worth pulling into mock analysis specifically because a low mood rating on a study session preceding a mock is a plausible, checkable explanation for a dip, in a way that re-reading solved questions never surfaces. If a bad mock sits after a run of low-mood sessions, that's worth naming in the analysis as an alternative to a content gap - and it's a hypothesis you can only form if the mood data exists to check against, which is the actual argument for logging it at all rather than just the score.

What to skip

Don't re-solve every question you already got right - that time belongs on the wrong and skipped ones. And don't spend the session comparing your percentile against anyone else's: as the data in percentile vs score shows, the same raw score has landed at percentiles more than 25 points apart across different papers in my own log, which makes a cross-person percentile comparison close to meaningless.

Making this repeatable past mock fifteen

A framework only works if it actually runs after every mock, which gets harder to sustain by hand once you're past a dozen. Karma Yogi lets you log a mock, then log each review pass against it tagged R1 through R7, so the trend-spotting in Step 5 and the topic comparison in Step 3 are sitting in the Insights dashboard rather than scattered across separate notebook pages. Start logging your mock analysis free.