Mock Test Analytics Compared: What a Percentile Actually Tells You on Each Platform

You finish a mock and the report says "92nd percentile, All India Rank 1,240". It feels precise. It is precise, in the sense that the arithmetic is correct. Whether it tells you anything about your chances depends on three things the report rarely shows: who you were compared with, whether they answered the same questions as you, and how many of them there were.

This post is about those three things. It is not a ranking of platforms. We run one ourselves (Pareeksha), and the honest way to talk about analytics is to explain the method, so you can check any platform's report, ours included. If you want the basics of reading a rank first, start with All-India Ranking Explained.

What a percentile is, in one line

Your percentile is the share of the comparison group that scored below you.

If 5,000 students are in the group and you are at rank 500, then 4,500 scored below you, and your percentile is 4,500 ÷ 5,000 × 100 = 90. (Platforms differ slightly in how they treat ties and whether they count you, which can move the number by a fraction. Ignore that; it does not matter.)

The formula is simple. Everything interesting is hidden in the words "the comparison group".

Question 1: Did everyone answer the same questions?

There are two common ways to build a mock test.

Fixed paper. One set of questions is written, and every student in the cohort gets that exact set. The real SSC and RRB exams work this way within a shift.

Drawn from a pool. The platform keeps a large bank of questions and assembles each student's test from it, sometimes by topic and difficulty tag. Two students "taking the same mock" may see different questions.

Pools have real advantages. You can retake a test and see new questions, and the platform can generate many tests from one bank. The trade-off is in the percentile.

Here is why. Say a pool tags questions as easy, medium and hard, and tries to give everyone the same mix. Tags are judgements made in advance. In practice, two "medium" questions can differ a lot in how many students get them right. If student A's draw happens to include several questions that turn out easier than tagged, and student B's includes several that turn out harder, A can outrank B with the same ability. The percentile then mixes your skill with the luck of the draw.

A small worked example. Suppose two students of identical ability each answer 100 questions. On A's draw, the questions are, on average, slightly easier, worth an extra 4 marks to a typical student. A scores 148 and B scores 144. In a crowded part of the distribution, 4 marks can be hundreds of ranks.

On a fixed paper this cannot happen, because there is no draw. A and B face the same 100 questions, so the only thing that separates them is how they answered.

We are not saying any particular platform randomises, or how it adjusts for it if it does. Some platforms may use fixed papers for some tests and pools for others, and some may correct for difficulty statistically. We could not verify the internals of other platforms' systems, so we will not guess. What you can do is ask: "In this ranking, did everyone see the same questions?" A good platform will be able to tell you.

On Pareeksha, the answer is yes by design. Every student in a cohort sits the same fixed paper, so "90th percentile" means "you beat 90% of the people who answered exactly your questions". That is the choice we made, and it has costs too, which we come to below.

Question 2: How big is the group?

A percentile from 50 students and a percentile from 50,000 students are not equally trustworthy.

Think of it this way. With 50 students, moving from 40th place to 35th place changes your percentile by 10 points (from 20th to 30th percentile). Five people having a good or bad day can swing your number a lot. With 50,000 students, you would need to pass 5,000 people to move 10 points.

There is also the question of how stable the group's shape is. In a small group, a handful of very strong or very weak students can stretch the distribution. In a large group, those outliers barely matter.

A rough guide, not a rule:

Group size on that exact paper	How to read your percentile
Under about 100	Treat as a rough indication only; watch your raw score and accuracy instead
A few hundred	Meaningful in broad bands (top 10%, top 25%, middle)
Several thousand or more	Fine-grained; differences of a few percentile points start to mean something

This cuts both ways for us. A fixed-paper cohort on Pareeksha is fair, but on a given paper it will often be smaller than the crowd on a very large platform. A fair comparison with fewer people is noisier than a fair comparison with many. If your report shows a small cohort, read the percentile in bands rather than to the decimal.

Large scheduled live tests are one way platforms get big groups on a common paper at the same time. Oliveboard, for example, advertised free SSC CGL Tier 1 live tests every Friday to Sunday with all-India ranking when we checked its site on 4 October 2026. Whether everyone in such a test gets identical questions is something to confirm with the platform.

Question 3: Who is in the group?

Even with the same questions and a big group, the people in the group shape your percentile.

• Serious and casual attempts mixed. If many students start a mock and submit after ten minutes, they fill the bottom of the list. Your percentile goes up without your preparation changing. Some platforms filter these out; many reports do not say.
• Early vs late attempts. Students who take a mock on release day are often the most regular ones. If you take it a week later, the group you are compared with may include more students who are revising or browsing.
• Different exams in one ranking. A general awareness test taken by both SSC and banking aspirants compares you with students who prepared for different syllabi.
• Re-attempts. If a platform allows unlimited re-attempts, check whether second attempts are counted in the rank. A student who has seen the paper before is not a fair comparison.

None of this makes percentiles useless. It means a percentile is a measurement of you relative to a specific crowd, and the crowd differs from platform to platform and test to test.

What "All India Rank" means when cohorts differ

"All India Rank" sounds like your position among all candidates for the exam. On a mock platform it means your position among the students who took that mock on that platform. Those are very different groups.

The real exam has a much larger pool than any mock, and it includes students who never take online mocks at all. So a mock AIR of 1,240 does not predict an exam rank of 1,240.

It also means two mock AIRs on different platforms cannot be compared. A rank of 1,240 out of 40,000 on one platform and a rank of 300 out of 3,000 on another work out to roughly the 97th and 90th percentile, but they come from different papers and different crowds. Neither is "your real rank". This is also why raw scores do not transfer: a 150 on one platform's paper is not a 150 on another's.

How to use analytics well, on any platform

1. Track your own trend on one series. The most reliable signal is your percentile and accuracy over 8 to 10 mocks on the same platform, not any single number. One test is noise; a trend is information.
2. Read percentile in bands. Top 10%, top 25%, middle half. Do not celebrate or panic over a 3-point move.
3. Check the cohort size if the report shows it. If it does not, ask.
4. Look at section-level numbers. Accuracy and time per section usually tell you more about what to study next than the overall rank does. Our guide to the analytics dashboard walks through ours.
5. Calibrate against official papers. Take an official previous year paper under timed conditions now and then. Compare your score with the published cut-off for that year. That is the one comparison where the questions are real. See our SSC CGL cut-off analysis for how to set a target from it.

A quick comparison of approaches

Approach	Strength	Watch out for
Fixed paper, everyone same questions	Percentile reflects only how you answered	Smaller cohort on a given paper; cannot retake with fresh questions
Questions drawn from a pool	Many fresh tests; retakes feel new	Percentile can mix ability with difficulty of the draw unless corrected
Large scheduled live test	Big group at the same time	Confirm whether questions are identical for all and how late attempts are ranked

FAQ

Is a 90th percentile on one platform the same as on another? No. Different papers, different crowds. Compare your own trend within a platform.

What cohort size makes a percentile reliable? There is no single cut-off, but a few hundred students on the same paper gives a usable broad picture, and several thousand lets small differences start to mean something.

Does my mock All India Rank predict my exam rank? No. The exam pool is far larger and different. Use mock percentile trends and official previous year papers to judge whether you are near the cut-off.

How does Pareeksha compute percentile? Everyone in a cohort sits the same fixed paper, and your percentile is the share of that cohort who scored below you. Full mocks are part of Pareeksha Pro; see the pricing page or try a mock.