Mock Test Analysis Guide: A 90-Minute Review, Error Log and Percentile Bands
Review a mock in 90 minutes: an error-log template, the accuracy versus attempts trade-off under +3/-1 marking, and what a percentile band means.
12 min read · updated 2026-11
Why the review matters more than the mock
A mock measures you. The review trains you. Two aspirants who score the same can improve at very different rates, because one reads the score and the other finds the five habits that cost the marks.
Take full mocks under real conditions. CAT 2026 is a 120-minute test with three sections of exactly 40 minutes each, no switching between sections, 3 marks for a correct answer and -1 for a wrong MCQ. Non-MCQ questions have no negative marking, and skipped questions score zero. Set your mock to the same rules, or the numbers will mislead you.
The 90-minute review
| Minutes | Step | What you produce |
|---|---|---|
| 10 | Scorecard | Per section: attempted, correct, wrong, skipped, minutes spent |
| 30 | Every wrong answer | Re-solve it, then label the error type |
| 15 | Skipped but solvable | Find out why you skipped; solve it without a timer |
| 15 | Right but slow | Find the faster method; note it in the log |
| 10 | Re-solve the three costliest questions cold | Check whether you now get them in under two minutes |
| 10 | Error log and next week's actions | Three actions, each tied to a topic or habit |
That adds to 90 minutes. Review the right answers you guessed as well; a lucky guess is a hidden error.
Accuracy versus attempts
- A: wrong = 6; score = 3 × 14 - 6 = 36; accuracy = 14/20 = 70%.
- B: wrong = 2; score = 3 × 12 - 2 = 34; accuracy = 12/14 = 85.7%.
- C: wrong = 11; score = 3 × 15 - 11 = 34; accuracy = 15/26 = 57.7%.
Answer: A scores 36. B has the best accuracy and C the most attempts, but both score 34. Marks peak between the extremes.
- Use 4p - 1.
- p = 1/4 (a pure guess among four options, as an illustration): 4 × 0.25 - 1 = 0.
- p = 1/3 (one of the four ruled out): 4/3 - 1 = +0.33.
- p = 1/2 (two of the four ruled out): 4 × 0.5 - 1 = +1.
Answer: Guess only after you can honestly remove at least one option. For non-MCQ questions there is no penalty, so always answer.
- Time share = 14 / 40 = 35%.
- Capping each at 2 minutes would have freed 14 - 6 = 8 minutes.
- At about 2.5 minutes a question, 8 minutes buys 3 more attempts (8 / 2.5 = 3.2).
- At 70% accuracy, each attempt is worth 3 × 0.7 - 0.3 = 1.8 marks, so 3 attempts add about 5.4 marks.
Answer: 35% of the section's time; about 5 marks on offer from a stricter time ceiling.
The error log template
Keep one sheet for all mocks. One row per question you got wrong, skipped or solved slowly.
| Mock and question | Topic | Result, minutes | Error type | Fix | Retest |
|---|---|---|---|---|---|
| M7 QA 12 | Remainders | Wrong, 3.5 | Concept | Re-learn cyclicity of powers | +3 days |
| M7 DILR 2 | Tournament set | Skipped, 0 | Selection | Rank sets by anchors before choosing | +10 days |
| M7 VARC 9 | Inference | Wrong, 2 | Misread | Underline the qualifier in the stem | +3 days |
- Concept: you did not know the idea. Formula: you knew it but could not recall it.
- Calculation: slip in arithmetic. Misread: you answered a different question.
- Time sink: right method, too slow. Selection: you chose the wrong question or set.
- Guess: you attempted with no basis.
Retest each logged question after three days and again after ten. If you miss it a second time, the fix was wrong. Rewrite it.
Reading the numbers across mocks
Do not chase one number. Each week, pick a single metric for each section, such as accuracy above 75%, no question over 2.5 minutes, or a minimum of two sets solved in DILR. Track it across three mocks before you change your strategy.
- Low attempts, high accuracy: you are too careful. Lower your ceiling per question and attempt the easy ones faster.
- High attempts, low accuracy: you are guessing and rushing. Slow down on selection.
- Good scores in one section only: move practice hours to the weaker section.
What a percentile band means
A percentile shows the share of test-takers who scored at or below you. It is relative, not absolute. A rough conversion is percentile = (N - rank) / N × 100, though exam bodies may treat ties differently.
- (100000 - 4000) / 100000 × 100 = 96.
- A 3-point band is 3% of 100000 = 3,000 candidates.
Answer: About the 96th percentile. A 3-point band holds about 3,000 candidates, so a few marks can move you across it.
- Average = (88 + 91 + 90 + 94) / 4 = 363 / 4 = 90.75.
- Range = 94 - 88 = 6 points.
Answer: Plan around a band of roughly 88 to 94, centred near 91, not around the best score.
- A mock percentile compares you with that platform's users, who may differ from the real exam pool. Use your own trend, not the absolute figure.
- Treat a predicted percentile as a band, never a point.
- Each institute sets its own cutoffs and weights, so check the official notice. CAT 2026 results are likely in the first week of January 2027, per the official bulletin.
The weekly cycle
- ☐ Take the mock under real timing and marking, with no pauses.
- ☐ Review it within 24 hours, using the 90-minute routine.
- ☐ Update the error log and pick three actions for the week.
- ☐ Retest the logged questions that are due, at three and ten days.
- ☐ Compare this mock with the last three on one metric per section.
- ☐ Change your strategy only after three mocks show the same pattern.
Common questions
How should I analyse a CAT mock test?
Give it 90 minutes within 24 hours: scorecard, every wrong answer, skipped but solvable questions, right but slow ones, then an error log with three actions for the week.
What is a good accuracy percentage in a mock?
There is no official benchmark. Under +3/-1 marking, you need over 25% accuracy on an attempt to break even; aim to beat that by a wide margin on every attempt you choose.
How many questions should I attempt in a section?
The official CAT 2026 pages do not state the number of questions. Attempt the ones you can solve inside your time ceiling and skip or guess the rest only if you can rule out an option.
What does a percentile band of 95 to 98 mean?
It means you scored better than roughly 95 to 98 percent of the test-takers in that pool. Treat it as a range, since small score changes move you inside it.