How Grade Boundaries Are Set — the UK vs the Rest of the World

In England, Wales and Northern Ireland, the raw mark a student needs for each GCSE or A-level grade is not fixed in advance — it is decided AFTER the exam has been sat, once senior examiners can see how the actual cohort performed on that year's actual paper. Combined with a statistical approach called comparable outcomes, this is why the mark needed for, say, a grade 7 can move up or down from one year to the next even though the exam looks broadly similar. That is a genuinely unusual design choice compared with much of the world, where a pass mark is often fixed before the exam is even sat.

Few parts of the UK exam system confuse parents more than grade boundaries. It feels intuitive that a grade should require a fixed number of marks — say, 70% for a grade 7 — every year. It doesn't work that way, and understanding why requires understanding what the system is actually trying to hold constant: not the mark, but the standard.

What actually happens after the exam

Once a paper has been sat and marked, the awarding body runs an "awarding" process before boundaries are finalised:

1. Statistical predictions. Using data about the national cohort — most significantly, students' prior attainment (for GCSEs, largely their primary-school Key Stage 2 results) — statisticians predict roughly what proportion of students should achieve each grade, based on how similar cohorts have performed on this qualification historically.
2. Senior examiner review. A panel of experienced examiners independently reviews real student scripts sitting right at the proposed boundary — the ones scoring just above and just below a candidate cut-off — and judges whether the quality of that work matches what a grade 7, or a grade 4, or a grade C, should look like.
3. Reconciliation. The final boundary is set where the statistical prediction and the examiners' qualitative judgement agree, or as close to that point as the evidence allows.

Ofqual's guidance on setting grade boundaries and its annual awarding documentation describe this process in detail, and it applies across AQA, Pearson Edexcel, OCR and WJEC/Eduqas under Ofqual's regulation.

Why this genuinely puzzles parents — and the logic behind it

The reason this system exists is called comparable outcomes, a principle Ofqual has used since 2011. The idea is that a national cohort's overall grade profile in a subject should reflect what you'd predict from that cohort's ability — measured through prior attainment data — rather than swing wildly because one year's paper happened to be harder or easier to write, or because teaching to a new specification hit unexpected snags.

Put simply: if this year's Year 11s are, on paper, about as strong as last year's Year 11s were, roughly the same proportion should come out with each grade — even if this year's exam questions turned out trickier. To make that true, the mark needed for each grade has to flex to absorb the paper's difficulty, because the proportion of students hitting the standard is the thing being held steady, not the raw score.

This is genuinely different from how most people intuitively think grading works, and it's exactly why a school newsletter reporting "boundaries dropped this year" isn't necessarily bad news — it often means the paper was judged harder, and the system compensated so the grade still represents the same standard of understanding as before. Ofqual's own explainer on comparable outcomes sets out this reasoning directly.

How other countries do it differently

Fixed-mark systems

Many national systems set a pass or grade threshold before the exam and do not adjust it afterwards. A student needing, for example, an average of 10 out of 20 to pass a qualification knows that number going in, and it does not move regardless of how the cohort as a whole performs that year. This trades away the UK's year-on-year standard-holding for simplicity and predictability — students and teachers always know exactly what score is needed, at the cost of the pass rate itself varying more from year to year as papers vary in difficulty.

Norm-referenced systems

Some systems go further and rank students explicitly against their own cohort rather than against any fixed mark at all. South Korea's national university entrance exam, the CSAT (Suneung), is a well-known example: results are reported partly as standard scores and percentile rankings relative to that year's test-takers, administered by the government-run Korea Institute for Curriculum and Evaluation (KICE). In a purely norm-referenced system, there is, by definition, always a top percentile and a bottom percentile — the comparison is to other students sitting the same exam in the same year, not to an externally-defined standard of knowledge.

A concrete fixed-mark example: the French baccalauréat

France's baccalauréat is the clearest real-world illustration of a fixed-mark system. Each subject is marked out of 20, the overall bac average needed to pass is set at 10 out of 20, and that threshold is published policy rather than something decided after the fact by examiners reviewing scripts. A student who averages 9.8 across their subjects has not passed, in a way that a GCSE student a fraction of a mark below a boundary sometimes has, once examiner judgement and comparable outcomes have been applied. France's Ministry of Education administers the bac nationally, and the fixed threshold is one reason French parents and pupils generally describe the system as more predictable, if less forgiving of an unusually hard paper, than the English one.

A hybrid worth knowing: Finland's matriculation examination

Finland's national matriculation exam (ylioppilastutkinto) sits closer to England's approach than either the French or Korean examples, without being identical to it. Grades — from the top laudatur down through six lower bands — are set using score distributions calculated across all candidates sitting a given subject exam nationally that year, a process run by the Finnish Matriculation Examination Board. Because the distribution is recalculated each sitting, the raw score needed for the top grade moves from year to year in a broadly similar spirit to English comparable outcomes, even though the mechanics of the calculation and the exam's role within a very differently structured school system are not the same.

Where the UK sits between the two

England's comparable-outcomes approach is best understood as sitting between these two poles: it isn't a fixed mark, because the actual number needed for a grade genuinely moves; but it isn't purely norm-referenced either, because the target is a standard of work (checked by real examiners reading real scripts), anchored statistically to what a similar cohort's prior attainment would predict — not simply "the top X% get an A."

Four systems side by side

Why this matters for a parent, not just in theory

The practical upshot for a UK parent is that a grade boundary moving down a few marks from the previous year is evidence the system is working as designed, not evidence of falling standards or an easier paper let through — it is very often the opposite. Conversely, a country using a fixed threshold like France's bac gives a family total clarity on what score is needed months in advance, but offers no correction if a particular year's paper turns out to be unusually difficult to finish in the time allowed. Neither trade-off is free; which one a family finds more reassuring is largely a matter of what kind of uncertainty — an unpredictable number, or an unadjusted paper — feels more manageable.

The takeaway

There is no single "correct" way to turn an exam script into a grade. England's system prioritises comparability of standard across years, at the cost of a boundary that moves and can seem opaque from the outside. Fixed-mark systems prioritise transparency and predictability, at the cost of the overall pass rate drifting with paper difficulty. Norm-referenced systems make relative ranking explicit rather than incidental. All three are legitimate, well-established approaches to the same underlying problem — deciding, fairly, where "good enough" actually sits.