Methodology

LEO (Longitudinal Education Outcomes)

HMRC-linked median earnings by qualification at 1, 3, 5 and 10 years after graduation. We ingest published DfE tables and store source_year and ingested_at on each row. Suppression rules from the publisher are respected — we do not impute suppressed cells.

Current UI rows: seeded illustrations until the read path serves ingested Supabase data on every course page.

Graduate Outcomes (HESA)

Employment, further study and work locations around 15 months after graduation. Public tables are irregular across releases — we document the file version in ingest metadata when live.

National Student Survey (OfS)

Student satisfaction fields are self-reported and can diverge from earnings outcomes. Where we publish NSS-derived figures we will show response rates alongside.

Subject code mapping

LEO, HESA and NSS use different subject hierarchies. We map to internal subject_canonical slugs (see lib/subject-mapping.ts) and document judgement calls here when a published code spans multiple teaching programmes.

Employability Index (Ewan+)

Separate from the Outcome Score above: an earnings-free composite used on the Compare courses screen for subscribers. When present, Graduate Outcomes employment (~15 months) is weighted 45%, NSS satisfaction 35%, and HESA continuation 20%. Missing fields are excluded and remaining weights renormalised — the same approach as the Outcome Score, without LEO earnings.

Outcome Score (MVP)

Temporary composite: blend of employment at ~15 months, NSS, continuation, and 5-year earnings normalised against tariff. Weights will be published as a table once seeding ends. We do not tune this score to flatter any sponsor — there are no sponsors.

Apprenticeship comparator

Net worth at 30 is a deliberately crude teaching model: constant tax haircut on salary, illustrative debt principal for the university route, and a simple growth curve for salaries. Real life includes rent, stock, bonuses, career breaks and tax bands — those are listed as limitations on the result page.

Scheme cards cite employer careers pages and Find an Apprenticeship; you must verify live adverts before applying.

AI (Claude)

Anthropic is used for narrative only. Numeric claims in Parent Reports and comparators must come from structured JSON we pass in — the model is instructed not to invent statistics.

Independence

Ask Ewan does not take payment from universities or employers to rank courses. Paid products (Parent Report, Clearing Intensive) are the revenue model; core comparison pages remain free to read.