Version 1.0 — published May 2026

MOT Checkup methodology — how we source, score and explain MOT data (v1.0)

This page is the source of truth for how every number on MOT Checkup is produced. It covers the data feeds that come in, the scoring logic we apply, what we deliberately exclude, and how often everything refreshes. Research posts and programmatic pages cite this version in their footer so readers can audit the inputs.

1. Data sources

Everything starts with publicly published government data. We do not scrape competitor sites and we do not buy data from undisclosed third parties.

DVSA MOT history API. The primary input. Provides every test result, mileage reading, advisory and defect for vehicles tested in Great Britain since 2005.
DVLA Vehicle Enquiry Service (VES). Tax status, SORN flag, CO2 emissions, fuel type, Euro standard.
Anonymised public records. For aggregate research posts (e.g. pass rates by region or make), we work from anonymised, vehicle-level DVSA records — never with personally identifiable owner data.
Named consumer reliability surveys. Where a research post triangulates owner-reported reliability with MOT data, the survey is named in-line and on the post itself, with year and sample size where the publisher has disclosed it.

MOT Checkup is not affiliated with the DVSA or DVLA. Where data is attributed it is attributed verbatim to the publishing body.

2. How we calculate reliability scores

A reliability score on MOT Checkup is a 0-100 figure that combines MOT-derived signals with named consumer-survey rankings. The recipe is intentionally boring: each input is normalised, weighted, and summed.

The headline inputs in v1.0 are the first-time MOT pass rate for the make/model, the rate of dangerous and major defects relative to the model average, mileage-anomaly frequency, and the cross-reference of named owner-survey ranks. None of these inputs alone is decisive — the score is a triangulation.

The first published research post that uses this score is Most Reliable Cars in the UK 2026. Its methodology box names the surveys, the weights, and the handling of small-sample models. Future research posts that re-use this score cross-reference the same box and any changes will bump this page to v1.1.

3. How we score common faults

Programmatic common-faults pages summarise the most frequent advisories and failure reasons for a given make/model. The score for each fault is a function of two things:

Frequency. How often the advisory or defect text appears in DVSA records for the model, normalised by the size of the sample.
Severity. The DVSA defect category — ADVISORY, MINOR, MAJOR or DANGEROUS — with DANGEROUS and MAJOR defects weighted more heavily than ADVISORY items.

Advisory descriptions are clustered: free-text variations of the same underlying issue (e.g. several ways of writing "front brake disc worn") are grouped before counting. We disclose the top clusters on the page so readers can sanity-check the grouping themselves.

4. What we exclude (and why)

There are categories of data we do not surface, on principle or for licensing reasons. Being explicit about these matters more than looking comprehensive.

Insurance write-off categories (Cat A/B/S/N). This data sits behind HPI/Experian licensing. We do not invent it, estimate it, or scrape it from secondary sources.
Outstanding finance markers. Same: a paid HPI/Experian database. A paid HPI check is the appropriate tool and we say so.
Owner identity. The DVSA's MOT records do not include owner names, addresses or contact data, and we do not attempt to resolve registrations to people.
Predicted resale value. Vehicle valuations require live trade-price feeds we do not currently license; the MOT check stops at history and condition signals.

5. Update cadence

MOT history per registration. Pulled live on each lookup from the DVSA API, with a brief cache to keep the page fast and respect DVSA rate limits.
Reliability scores and common-faults aggregates. Recomputed weekly from the latest DVSA records. The page shows the date of the most recent recompute.
Research posts. Published with a stamped data cut-off and refreshed annually unless a major DVSA dataset release warrants a sooner update.
This methodology page. Versioned. Cosmetic updates do not bump the version; substantive changes to inputs, weights, or exclusions do.

6. Changelog

v1.0 — May 2026. Initial public methodology. Establishes data sources, reliability score recipe, common-faults scoring, exclusions and cadence.

Frequently asked questions

Why do you version the methodology?

Reliability scoring, advisory weighting and source mix all change over time as we add data and improve the model. Versioning lets us state, on every research post and report, exactly which methodology produced the numbers. If we change the maths, we publish a new version and keep the old one accessible — readers should never wonder which formula they're looking at.

What changes will trigger v1.1?

Any of the following: a new data source added to the input mix (e.g. a fresh consumer survey we license), a change to the reliability score weighting, a change to how we cluster advisory descriptions, or a change to the exclusion list. Cosmetic copy changes do not bump the version.

Where do you publish methodology updates?

On this page. Every version is dated. Research posts cite the methodology version they were generated under in their "About this research" footer, so a 2026 post will keep referring to v1.0 even after we publish v1.1 in future.

Can I see the underlying DVSA data yourself?

Yes. The DVSA publishes the MOT history dataset openly, and you can query it via the official MOT history API. We do not paywall the raw inputs — the value we add is the layering, the chart, the common-faults summary and the AI explanation.

Why don't you license HPI/Experian data?

Cost and product focus. HPI and Experian licence finance and write-off data on a per-check basis at prices that would force us to charge for the core MOT check. We have chosen to keep the MOT check free and signpost paid HPI checks for buyers who need that depth.