A mark a minute holds. The question quadruples.

The pace barely moves

Marks per minute is the closest thing this dataset has to a measure of intended working speed. It sits at 0.80 at KS1 and 0.78 at A Level, having wandered up to 1.07 at 13+ and down to 0.68 at IGCSE in between. There is no trend across the ladder worth naming.

That is a mildly surprising result if you expect senior material to be paced harder. It is much less surprising once you notice that mark allocation and time allocation are set together by the same person against the same conventions, and that the convention itself is broadly stable across UK specifications. The rule of thumb survives because it is baked into how the material is written, at every level.

The two visible outliers both have plain explanations. The 11+ and 13+ figures run high because that material is written as short, timed, high-pressure drills where a mark is meant to be earned quickly. The IGCSE figure runs low on one of the smallest groups in the corpus.

The question gets much longer

The same table read down its first two columns tells a different story. Marks per question rises from 1.52 at KS1 to 3.62 at GCSE to 5.82 at A Level, and target minutes per question rises alongside it from 1.91 to 4.35 to 7.51.

An A Level question is therefore about 3.8 times the size of a KS1 question, and takes about 3.9 times as long, while the pace at which marks arrive is unchanged. Whatever else is happening between key stages, a student is being asked to hold one line of reasoning together for four times as long before anything is scored.

This is a useful reframing for anyone planning revision. A student who can work accurately for two minutes at a time is well equipped for KS1 material and badly equipped for A Level material, even with identical subject knowledge, and no amount of extra short-question practice fixes that. Practising the sustained answer is a different exercise from practising the topic, which is why our own worksheet packs print full mark schemes rather than answer keys.

Where the practice actually sits

The corpus is heavily weighted towards GCSE, which carries 1,413 of the 2,899 packs and 21,135 questions on its own. That reflects where demand is rather than any judgement about where practice matters most.

Totalled across every level, the material represents about 3,203 hours of intended working time. That number is worth treating carefully: it is the time to work through every published pack once, not a curriculum, and no student would sensibly attempt more than a small fraction of it. It is most useful as a sense of scale for the corpus, and as a denominator when comparing subject groups against each other.

A calculator is permitted on 39.2 per cent of packs, which is a reasonable proxy for how much of the corpus is non-calculator numerical work and how much is written or reasoning-led.

By subject group

The same totals broken out by subject group, ordered by question count. This is the level most planning actually happens at, and it is the table most likely to be useful to someone building a scheme of work or a revision timetable.

If you want to turn a mark total and a time limit into per-question pacing for a specific paper, the exam pacing calculator does that arithmetic directly, using the same marks-per-minute relationship measured here.

Download the data

The full dataset is published as a CSV and as JSON, one row per pack, with the level, subject group, topic, question count, mark total, target minutes, calculator rule and page counts for every one of the 2,899 packs. It is generated from the same source the figures above are computed from, so the file and the page cannot disagree.

It is released under CC BY 4.0. You may republish it, chart it, load it into a spreadsheet, or build something else on top of it, including commercially, as long as Revision Library is credited and the licence is linked. If you do something interesting with it we would like to see it.

Method

The dataset is every worksheet pack published on this site: 2,899 packs across 39 subject groups and 9 levels, totalling 52,035 questions, 163,755 marks and 27,623 printed pages. Every pack was written and reviewed in-house, so the fields below exist for all of them rather than for a subset.

Each pack records four things at the moment it is authored: the number of questions, the total marks, a target working time in minutes, and whether a calculator is permitted. Marks are allocated to match the mark-scheme conventions of the specification the pack is written against. Target minutes are the time the author intends a student working at the expected standard to take, set alongside the mark allocation rather than derived from it.

The three ratios in this study are simple quotients of those fields, summed over a group before dividing rather than averaged per pack, so a long pack counts for more than a short one. Marks per question is total marks over total questions; target minutes per question is total minutes over total questions; marks per minute is total marks over total minutes.

Levels are the level label each pack was published under. A pack appears in exactly one level and one subject group, so no row is double counted. The full row-level data, one line per pack, is published below under a licence that allows reuse.

Questions

Is the one-mark-a-minute rule actually true?

In this corpus it is a good approximation at every level, with the pace ranging from 0.68 to 1.07 marks per minute. It is a rule about how material is written rather than a law about how students work, so treat it as a planning heuristic and not as a target to hit.

Does this prove A Level is four times harder than KS1?

No, and the data cannot support that claim. It shows that one A Level question carries about 3.8 times the marks of one KS1 question and takes about 3.9 times as long. Difficulty involves content, abstraction and prior knowledge, none of which this dataset measures.

Why is the IGCSE row so different from the GCSE row?

Mostly sample size. IGCSE is the smallest group in the corpus by some distance, so its ratios are unstable and move noticeably when a handful of packs are published. Read the large rows with confidence and the small ones as indicative.

Can I use this data in my own article or research?

Yes. The dataset is published under CC BY 4.0, so reuse of any kind is fine as long as Revision Library is credited and the licence is linked. The citation line on this page is written to be copied directly.

How often is this updated?

The figures are recomputed from the corpus every time the site is built, so publishing new packs moves the numbers on this page and the rows in the downloadable file together. The review date at the top records the last time the study's wording was checked against them.