AQA A-Level Maths (7357): The Large Data Set Discrepancy

AQA A-Level Mathematics is specification 7357, Version 1.3, dated 31 January 2018, regulated by Ofqual under qualification number 603/1164/2 and first taught from September 2017. Its compulsory large data set is still described in that specification as a 2014 DEFRA food-spending extract — yet the file AQA actually serves, and tested in its June 2025 Paper 3 exam, is an unrelated 3,827-row vehicle emissions dataset.

Which edition of AQA's specification is this, and is it still current?

This page describes AQA A-Level Mathematics 7357, Version 1.3, dated 31 January 2018 — the specification document AQA's own filestore still serves as the current PDF for the qualification, with no later issue number published anywhere on it or on AQA's live specification-at-a-glance page, which uses identical section numbering and text. AQA does not date-stamp a superseding edition; the version and date printed on the cover are the only ones in circulation.

Ofqual's own qualifications register independently verifies the code: entering 603/1164/2 returns “AQA Level 3 Advanced GCE in Mathematics … Qualification number 603/1164/2 … Availability Status Available to learners National availability England, Northern Ireland, International.” Wales is not named in that field. AQA's own specification and website say nothing about country availability one way or the other, so this is the regulator's own classification, not AQA's — and it states only what is quoted here, not that no Welsh school could ever offer the qualification through some other arrangement.

The same register entry gives a regulation start date of 27 February 2017 and an operational start date of 29 August 2017 — independent corroboration, from Ofqual rather than from AQA's own specification cover, that AQA A-Level Mathematics was first taught from September 2017. Maths and Further Maths decoupled from linear AS assessment in 2017; several science A-levels decoupled two years earlier, in 2015, and the two dates should never be read across to each other.

Does AQA's specification correctly describe the large data set it actually tests?

Section 3.21.1 of AQA's still-current specification is explicit: “The dataset featured in statistics questions for exams in 2018 and 2019 is an extract of a dataset that underpins DEFRA ‘Family Food 2014 report’ (published in 2015). The dataset contains information on purchased quantities of household food and drink by Government Office Region from 2001 until 2014. The specific extract that students need to be familiar with for the exam is only available via the AQA website.” That is the only description of the large data set anywhere in the document, and it has not been reissued or renumbered since 31 January 2018.

AQA's own assessment-resources search for 7357, filtered to “large data set”, lists the current file quite differently: “Assessment materials: large data set - for AS exams from 2019 and A-level exams from 2020 - Mathematics XLSX 427.76 KB Published: 02 Sep 2018.” It is the only large-data-set file in that search that is not gated behind a myAQA login — the files still titled for the “Family Food” report are all marked “Published: 01 Aug 2024”, a re-upload of the same old file rather than a new one. Downloading the ungated file — 438,030 bytes, a genuine Excel workbook, one sheet, 3,827 rows — shows its columns are Make, PropulsionTypeId, BodyTypeId, GovRegion, EngineSize, YearRegistered, Mass, CO2, CO and NOX: car registration and emissions data, not food spending. AQA's own field notes inside the file describe “the English region of the current registered keeper” and “the mass of vehicle in kg, including the mass (75kg) of the average driver.”

A real, recently released exam paper settles which dataset students are actually tested on. AQA's June 2025 Paper 3, footer G/Jun25/7357/3, question 20, reads: “The oxides of nitrogen emissions, F g/km, of a car registered in 2000 can be modelled by a normal distribution with mean 0.41 and standard deviation 0.07,” and part (e) asks students to explain why a researcher “cleaned the data” after collecting “data on the oxides of nitrogen emissions from the Large Data Set for cars registered in 2002 and 2016.” That is the vehicle dataset, examined by name as “the Large Data Set”, in the most recent released paper for this specification.

Three AQA-published documents do not agree. The specification names a 2014 food-spending extract and says it applies “for exams in 2018 and 2019”. AQA's own assessment-resources listing titles the actually-downloadable file for exams “from 2020” onward and shows it to be vehicle data. And a genuine 2025 exam paper tests that vehicle data under the name “the Large Data Set”. None of the three has been withdrawn or corrected as far as this research found — AQA's news, key-dates and assessment-resources pages for 7357 carry no erratum notice. A student revising directly from the specification's own section 3.21.1 would prepare to be familiar with household food-spending patterns by Government Office Region; the exam in front of them asks about car emissions by year of registration.

AQA's separate notes-and-guidance page for the large data set — a different document again from the specification — is consistent with the file AQA actually serves, not with section 3.21.1: it tells teachers the exams “will include questions or tasks that relate to the prescribed large data set, giving a material advantage to students who have studied it”, and recommends using it “as a classroom tool” with “basic knowledge of spreadsheet packages such as Microsoft Excel or Geogebra” rather than memorising it, because it is “too large to be taken into an exam”. That page does not mention DEFRA or food spending anywhere.

What does AQA choose for itself, and what does the DfE or Ofqual fix for every board?

Content is not AQA's decision. The Department for Education's own subject content document represents 100% of A-level Maths content, so 7357 could not diverge from Edexcel or OCR even if AQA wanted it to — our A-Level Maths guide sets out that fixed content and the resulting five-board comparison, so it is not repeated here. Two design choices genuinely are AQA's own.

The first is how AQA spreads its assessment objectives across its own three papers. Ofqual fixes AO1 at 50%, AO2 at 25% and AO3 at 25% for every board, but AQA's specification table, section 4.2.1, applies that split identically to Paper 1, Paper 2 and Paper 3 individually, not merely as a three-paper average:

Every column reads the same. AQA did not have to design it this way — Ofqual's rule only fixes the overall 50/25/25 split, not how a board divides it paper by paper — so a student cannot expect one AQA paper to lean more heavily on routine technique than another; the reasoning-and-modelling questions carrying AO2 and AO3 are spread evenly across all three sittings.

The second is which real-world dataset carries that statistics content — the large data set covered above, which AQA has chosen to make vehicle registration and emissions data rather than the sporting, environmental or other datasets used by other boards.

AQA also assigns its own DfE discount (classification) code to the qualification: RB31. The specification explains what sharing a code with another qualification would mean in practice — “If a student takes two specifications with the same discount code, further and higher education providers are likely to take the view that they have only achieved one of the two qualifications” — advice AQA prints as a caution to check before enrolment, not a live warning that 7357 currently shares a code with anything else.

Our wider A-Level tuition hub lists the other subjects we cover, and students moving on from AS Mathematics can see AQA's separate AS specification on our AS Mathematics page.

Frequently asked questions about AQA A-Level Maths