CAT Mock Test Analysis: The 45-Minute Routine That Moves Your Score
Taking a mock measures where you are. It does not move you. The improvement comes entirely from the review afterwards — and most candidates spend four minutes glancing at a score, then book the next mock.
That is why scores plateau. Below is a minute-by-minute routine, the four error types that each need a different fix, and how to tell which of your problems is actually costing you marks.
The ratio that decides everything
A CAT mock takes 120 minutes. Budget at least 45 for the review — roughly 40 percent of the attempt time. If you cannot afford that, take fewer mocks. Twenty analysed properly beats forty that were not, and it is not close.
Quick Answer (30-Second Read)
Time needed
45 minutes minimum
Same day, after a short break
The four buckets
Silly · Conceptual · Time · Guess
Each needs a different fix
- Break for 15 minutes first. Analysing while fried produces shallow review; leaving it to tomorrow loses why you chose each wrong option.
- Sort every error into four buckets before solving anything. Five minutes, four tallies.
- Read accuracy before score. The total hides both attempt count and accuracy inside one figure.
- Review the correct answers that took too long — those are time errors hiding inside right answers.
- Log the clue you missed, not the fact you were wrong. "Forgot to check the units" is usable; "got it wrong" is not.
Pattern references per the official CAT information bulletin at iimcat.ac.in.
CAT Mock Test — VARC, DILR & QA on the real interface
This routine needs time-per-question data and time-inside-each-DILR-set. A mock that only returns a score leaves you doing the diagnosis by hand.
The 45-minute routine
Take a fifteen-minute break first, away from the screen. Then do this the same day — by tomorrow you will have lost your memory of why you picked each wrong option, and that reasoning is the most valuable material in the whole exercise.
0–5 min · Sort, do not solve
Every wrong answer into one of four buckets: silly, conceptual, time, guess. Four tallies. No solving, no reading solutions.
5–12 min · Open the time view. Flag every question over 150 seconds and every DILR set over 8 minutes. One question each: should I have abandoned this?
12–17 min · The TITA and blank audit. Count TITA questions left blank. Count questions you never reached. Both are pure lost marks with no knowledge cause.
17–32 min · Re-solve the conceptual errors, closed book. Not reading the solution. Reading produces a warm feeling of understanding that evaporates under a clock.
32–40 min · Log each error with topic and the specific clue you missed.
40–45 min · Write one change. Specific enough to check next time. "Abandon a DILR set at 4 minutes with nothing locked" beats "manage time better".
If you only have fifteen minutes
Do the bucket sort and write the one change. Those two carry more of the value than any other pair, and they take twelve minutes between them. Skipping review entirely because you cannot do the full forty-five is how the habit dies.
The four buckets and their different fixes
This is the core of the method. Each bucket has a completely different remedy, and treating every error as "study more" is the main reason preparation stalls.
| Bucket | What happened | The fix | The wrong fix |
|---|---|---|---|
| Silly | You knew it. Misread the question, missed a "not", slipped a sign | A slower, deliberate reading pass. Underline what is actually asked | Studying the topic again — you already know it |
| Conceptual | You did not know the underlying idea | Go back to the topic properly, then drill it in a sectional | Memorising this one question's answer |
| Time | Could have solved it, but rushed — or overspent and had to hurry elsewhere | A firmer abandon rule with a per-question ceiling | Trying to "go faster" without changing what you skip |
| Guess | No real basis. You marked something anyway | Stricter attempt discipline — eliminate two options or leave it | More mocks. This is a decision habit, not a knowledge gap |
After five or six mocks, look at which bucket is consistently largest. That single fact tells you what the next fortnight should be, and it is usually not what candidates assume.
A candidate whose biggest bucket is silly does not need more study and will not improve by doing more. They need to read more carefully, which is a completely different intervention. A candidate whose biggest bucket is guess has no knowledge problem at all — they have an attempt-discipline problem, fixable inside a week.
Which bucket is actually costing you the most?
Most candidates guess wrong. The time-per-question and blank-TITA data turn a hunch into a number you can act on.
- Time per question
- Marks lost to negatives
- Blank TITA count
Reading the numbers correctly
Your total sits at the top of the report and is the least useful figure on the page, because it compresses two independent things — how much you attempted and how well — into one number that hides both.
| Metric | What it actually tells you | Act on it when |
|---|---|---|
| Sectional percentile | Which section is closest to failing a school's sectional screen | Any section trails the others by 10+ percentile points |
| Accuracy per section | Whether you are attempting past your reliable range | Below 70% anywhere |
| Time per question | Where the section actually went | Two or three questions exceed 150 seconds |
| Questions never reached | A pacing problem, not a knowledge one | More than four in any section |
| Blank TITA count | Free marks declined for no reason | Anything above zero |
| Marks lost to negatives | Whether guessing is disciplined or expensive | Losing more than 5 marks in a section |
Read accuracy before score, every time. A candidate attempting 16 questions at 88 percent accuracy is in far better shape than one attempting 22 at 64 percent, even when the totals look similar — because accuracy reflects how well you know the material while attempt count is a dial you can adjust in a week.
DILR needs its own review
The four buckets work for VARC and QA. DILR needs one extra question, because a DILR failure has two completely different possible causes.
| What the data shows | The diagnosis | What to practise |
|---|---|---|
| You picked hard sets and solved them slowly | Solving problem | More sets, focus on speed of arrangement |
| You picked hard sets while easier ones sat untouched | Selection problem | The prediction drill — rank sets before solving |
| You spent 11 minutes on one set and got nothing | Abandon-rule problem | Fixed 4-minute checkpoint, enforced |
| You solved two sets fully and stopped | Nothing wrong | That is a good DILR section |
Distinguishing rows one and two is the whole point, and you cannot do it without knowing how long you spent inside each set. Candidates routinely conclude "I am bad at DILR" when the data says they are fine at solving and poor at choosing — which needs a different fortnight of work entirely.
The last row deserves saying out loud. Two sets solved cleanly and the rest left untouched is a strong DILR performance, not a failure. Candidates who review that section looking for what went wrong sometimes talk themselves out of a strategy that was working.
An error log that is actually useful
Most error logs fail for one reason: they record that something was wrong rather than why. "Q34 — geometry — wrong" is a fact you cannot act on.
| Date | Section | Topic | Bucket | What I actually missed |
|---|---|---|---|---|
| 12 Aug | QA | Time & work | Silly | Read "in 6 days" as "for 6 days" |
| 12 Aug | VARC | RC inference | Conceptual | Chose an option true in the world but unsupported by the passage |
| 12 Aug | DILR | Arrangement | Time | Started the set with most variables; two easier ones untouched |
| 12 Aug | QA | Mensuration | Conceptual | Used slant height where vertical height was needed |
That last column is the entire point. Each entry names something checkable next time. After about thirty entries the log stops reading as a list of mistakes and starts reading as a list of two or three recurring blind spots — and those, not the syllabus, are what you should be drilling.
The three-strike rule
When a topic appears three times in your log, stop treating it as an error and start treating it as a project. Block two study sessions for it, then a sectional test on that topic alone. A topic that has beaten you three times will beat you in the exam unless something deliberately changes.
Why your mock score has stopped moving
A flat line across eight or ten mocks is the most common frustration in CAT preparation, and the instinct — take more mocks — is almost always wrong.
| What the plateau looks like | What is actually wrong | This week |
|---|---|---|
| Same score, same weak topics | Diagnosis never converted into study | Stop mocking. Two study days on your top log topic, then a sectional |
| Score flat, accuracy rising | You are improving — attempts are the lag | Nothing is wrong. Push attempt count by three per section |
| Score flat, accuracy falling | Chasing volume past your peak | Cut attempts by four. Rebuild accuracy, then rebuild volume |
| Score swinging wildly | Too few attempts for a trend to exist | Nothing yet. Below 15 mocks, variance is expected |
| One section stuck, others rising | You are practising what you enjoy | Sectionals only in the stuck section for two weeks |
The second row catches people out most. Rising accuracy with a flat score is progress that has not surfaced yet — you are getting more reliable at the same volume, and the marks arrive when you push attempts. Changing strategy in the middle of that is how candidates undo work that was about to pay.
The last row is the most common and least comfortable. Almost everyone practises the section they are already good at, because it feels better. Check your last month honestly: if your weakest section got the least practice time, the plateau is fully explained.
Key Takeaways
- Budget 45 minutes of review per 120-minute mock. If you cannot afford that, take fewer mocks.
- Sort errors into silly, conceptual, time and guess before solving anything — each needs a completely different fix.
- Read accuracy before score. 16 attempts at 88% beats 22 at 64%, even when the totals look alike.
- DILR needs an extra question — was this a solving problem or a selection problem? They need different fortnights.
- Log the clue you missed, not that you were wrong. Three appearances of a topic makes it a project.
- Flat score with rising accuracy is progress, not a plateau. Push attempts rather than changing strategy.
People also search for
How long should I spend analysing a CAT mock test?
At least 45 minutes for a 120-minute mock, which is roughly 40 percent of the attempt time. Early in preparation you may need longer, because there is more to learn from each paper. If you consistently spend under 15 minutes you are collecting scores rather than improving, and cutting your mock count to make room for proper review will raise your final percentile more than the extra papers would. Twenty mocks analysed properly beats forty that were not.
What are the four error buckets in mock analysis?
Silly errors, where you knew the concept but misread or slipped. Conceptual errors, where you did not know the underlying idea. Time errors, where you rushed something you could otherwise have solved, or overspent and had to hurry elsewhere. And guesses, where you had no basis and marked something anyway. Each needs a different remedy — silly errors need a slower reading pass, conceptual errors need study, time errors need a firmer abandon rule, and guesses need attempt discipline. Treating all four as "study harder" is why candidates plateau.
Why is my CAT mock score not improving?
Usually because the diagnosis from each mock is never converted into study. If your last eight reports have all flagged the same two weak topics and you have not spent dedicated days on them, a ninth mock will report the same thing. Check three other possibilities too: accuracy rising with a flat score is actually progress that has not surfaced yet; accuracy falling means you are chasing volume past your peak; and one section stuck while others rise usually means you are practising what you enjoy rather than what is weak.
Should I review questions I got right?
Only two kinds. Correct answers that took more than about 150 seconds are time errors hiding inside right answers, and they cost you the questions you never reached. And correct answers you guessed, because a lucky guess will not repeat and it inflates your accuracy figure misleadingly. Skip everything else — reviewing questions you answered quickly and confidently is the most comfortable part of analysis and the least useful.
How do I analyse the DILR section of a mock?
Ask one extra question beyond the four buckets: was this a solving problem or a selection problem? If you picked hard sets and solved them slowly, that is solving and needs more practice. If you picked hard sets while easier ones sat untouched, that is selection and needs the prediction drill instead. You cannot distinguish the two without knowing how long you spent inside each set, which is why DILR review depends on having that data. Two sets solved cleanly with the rest untouched is a good section, not a failure.
When should I analyse a mock — immediately or the next day?
Take a fifteen-minute break away from the screen, then analyse the same day. Reviewing immediately while mentally fatigued produces shallow work, and you are also still emotionally attached to the score, which distorts your judgement about which errors were genuinely careless. Leaving it until tomorrow is worse, because you lose your memory of why you chose each wrong option, and that reasoning is the single most valuable thing in the exercise.
What should a CAT error log contain?
Five fields per entry: date, section, topic, error bucket, and the specific clue you missed. The last one is what makes the log useful. Writing "got it wrong" teaches you nothing on review, while "chose an option true in the world but unsupported by the passage" gives you something to check for next time. After around thirty entries the log stops being a list of mistakes and becomes a list of your two or three recurring blind spots, which is what you should actually be drilling.
How many mocks before my score starts improving?
Expect noise rather than trend for the first ten to fifteen attempts, because paper difficulty varies and your own strategy is still changing. Real movement typically appears once you have both enough attempts for a trend and a review habit that converts each diagnosis into study. If you have taken twenty mocks with proper analysis and nothing has moved, the problem is almost certainly that the analysis is not being acted on — check whether your weakest section actually received the most practice time last month.
Should I re-solve every wrong question after a mock?
Only the conceptual ones, and closed-book rather than by reading the solution. Reading a worked solution creates a comfortable feeling of understanding that disappears the moment you meet the same question type under a clock; re-solving it yourself does not. Silly errors do not need re-solving — you already knew the method. Time errors need a decision about your abandon rule rather than another attempt. And guesses need attempt discipline, not more practice on that particular question.
Does analysing mocks actually improve percentile?
It is the part of mock practice that produces improvement. Taking a mock measures where you are; it does not change where you are. The change comes from the study that follows the diagnosis, which is why a candidate taking twenty well-analysed mocks reliably outperforms one taking forty unanalysed ones. It also explains the most common plateau in CAT preparation — candidates respond to a flat score by booking more mocks, which adds measurement and no treatment.
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.
