The first thing
The questions that carried no negative marking are not in the paper any more.
You have probably heard that decision-making questions carry no negative marking. That was true in 2011, 2012 and 2013. In 2014 both the questions and the exemption left the paper together.
| Year | Penalty applied to | The exempt questions were |
|---|---|---|
| 2011 | Every question except 73 to 80 | 8 decision-making items |
| 2012 | Every question except 74 to 80 | 7 decision-making items |
| 2013 | Every question except 75 to 80 | 6 decision-making items |
| 2014 – 2026 | Every question. No exemption. | None. The block was removed. |
From 2014 the line printed on the paper is plain: there will be penalty for wrong answers marked by the candidate. The exemption sentence is not shortened or moved. It is gone. You can check this yourself — download any paper below and open its instruction page.
The exemption was never given to a topic. It was given to one block of questions at the end of the paper. In 2011 that block held eight items and the exemption ran from 73 to 80. In 2012, seven items, 74 to 80. In 2013, six items, 75 to 80. The counts match in all three years.
Since 2014 the block is not in the paper at all. The line that opens it — “each item describes a situation and is followed by four possible responses” — does not appear once in any paper from 2014 to 2026. So two things are true together: guessing is no longer free anywhere, and practising decision-making questions is time spent on a question type that has not been asked for thirteen years.
The effect is direct. If you sit down believing the last few questions are free to guess at, you are losing one third of the marks for every wrong one and not noticing.
What has never changed
The frame has not changed. What sits inside it has.
Questions
Exactly eighty, every year since 2011.
Marks
Two and a half marks per question.
Hours
Ninety seconds per question, on average.
To qualify
33 per cent of 200. Going above it does nothing for your rank.
Marking more than one answer counts as wrong even if one of them is right. A question left blank carries no penalty. Both of those have been printed the same way for sixteen years.
One thing fewer people know: UPSC sometimes drops questions after the exam and scores out of what is left. The 2014 key shows 74 items and 185 marks. The 2024 key shows none dropped. So read that year’s key before adding up your marks.
What that means in practice. You need 66.67 out of 200 to qualify. Aim for 80. That is about five questions of cushion, so that two or three surprises on the day do not take you under.
Where the marks are
Two blocks decide this paper, and they are the same size.
This is CSAT 2024, placed question by question. It is the only paper where all eighty questions were located in the scan, so it is the only one that can carry a full breakdown.
Comprehension
Twenty-seven of eighty. These sit under a printed passage instruction, so there is no guesswork in counting them.
Basic numeracy
Exactly as many. Numbers, percentages, work, ratios, ages and algebra.
Logical reasoning
Coding, direction, series, clocks, calendars and conclusion questions.
Data interpretation
In this paper all ten are data sufficiency. Most people treat this as a small topic.
| Part of the syllabus | Questions | Share | What is in it |
|---|---|---|---|
| Comprehension | 27 | 33.8 % | Questions under a printed ‘read the following passage’ block |
| Basic numeracy | 27 | 33.8 % | Numbers, percentages, work, ratios, ages, algebra |
| Logical reasoning | 16 | 20.0 % | Coding, direction, series, clocks, calendars, conclusions |
| Data interpretation | 10 | 12.5 % | All ten are data sufficiency, which the syllabus places here |
The part that is most underrated
The syllabus reads: Data interpretation (charts, graphs, tables, data sufficiency etc.). Data sufficiency is officially part of it. In 2024 there were ten such questions — twelve and a half per cent of the paper. Preparation that writes data interpretation off as “two or three questions” is writing off all ten.
Comprehension in the other years
Between twenty-five and thirty questions, every year.
Counting these needs no reading at all. The paper prints how many items follow each passage, and that is the count used here. Only the years where the scan located at least 95 per cent of the questions are shown.
| Year | Comprehension | Of 80 |
|---|---|---|
| 2017 | 30 | 38 % |
| 2020 | 25 | 31 % |
| 2022 | 27 | 34 % |
| 2024 | 27 | 34 % |
What this means for you
Five things that follow directly from the numbers.
-
Reading and maths are level
In 2024 they were twenty-seven each. Most preparation runs as though the paper were a maths paper. Leaving either one aside means leaving aside a third of the paper.
-
Data sufficiency is not a small topic
Ten questions in 2024. The method takes one evening to learn — test each statement on its own, then the two together — and it applies to all ten.
-
The time spent on decision making can go somewhere else
Those questions have not been asked once since 2013 — thirteen papers, not one.
-
A block came back in 2026
Questions 73 to 77 were on communication and interpersonal skills — memoranda and minutes, barriers to communication, conflict at work, designing a questionnaire, circular letters. The first time this block had appeared since 2013. Twelve and a half marks.
-
A blank takes nothing from you
Every question has carried the penalty since 2014, but a question left blank costs nothing at all. A question you are unsure of is better left alone.
How this was worked out
The method, and where it stops.
All sixteen papers are scans with no text in them, so they were first read by machine in both scripts.
Comprehension is counted without reading a single question. Above each passage the paper prints how many items follow it, and that is the number counted. A faint scan cannot change it.
The 2024 breakdown was made by hand, all eighty questions placed one at a time. It is the only paper where every question was located. The four questions where the placing is a matter of judgement are written down separately, so that a disagreement can be about one question rather than about the total.
The other fifteen papers do not carry a full breakdown here, and that is deliberate. Their scans lost too many questions, and a percentage taken from an incomplete sample misleads.
This page changes when the numbers change. Every paper is free to download — count it yourself, and tell us if we have got something wrong.
Now find out where your marks are in all this.
Twelve questions, fifteen minutes, no charge. The result names the three topics with the most marks in them for you.