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.
• Bulk DVSA MOT records. The aggregate research posts on /research are computed from bulk DVSA MOT History records — one row per vehicle — obtained under our DVSA-approved MOT History API key. These records carry registration marks, so they are held internally, are never exposed anywhere on the site, and are aggregated before anything is published. Every published group is subject to a minimum sample size, set per table and ranging from 2,000 to 20,000 underlying records — vehicles, or defect lines where the table counts defects. Nothing we publish from this source — page, table, chart, CSV or embed — contains a registration mark or a VIN.
• DVSA/DfT anonymised MOT tests and results. The bulk open dataset published on data.gov.uk under the Open Government Licence. It carries no registration mark and no VIN; vehicles are linked by a derived Vehicle ID. This is the source of our model-level reliability benchmarks.
• What we do not do. The MOT record contains no owner name, address or contact detail, and we hold none. We do not attempt to resolve a registration to a person.
• 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.

1.1 Licensing and attribution

These sources do not share a licence, so we do not describe them as if they did. Where a licence exists we name it; where none has been published we say that instead of inventing one.

• Our own analysis. The tables, charts and CSV downloads we publish are released under CC BY 4.0. That grant covers MOT Checkup's aggregation, ranking and presentation. It does not extend to the underlying records or to anyone else's survey scores.
• DVSA MOT History API and bulk download. The underlying MOT data is Crown Copyright. DVSA publishes no data licence and no terms of use for this product, so we do not state one and we make no representation on DVSA's behalf. We describe the source instead: DVSA MOT History records, obtained under our DVSA-approved API key.
• DVSA/DfT anonymised MOT tests and results. Published on data.gov.uk under the Open Government Licence. Where a figure comes from that dataset we carry the attribution the licence itself specifies: "Contains public sector information licensed under the Open Government Licence."
• Named consumer reliability surveys. Scores from What Car?, Honest John, Warrantywise, Auto Express and JD Power remain the property of those publishers. We cite them, with the publisher named against each score; we do not license them onward, and re-users should consult the original publisher.

Corrected 28 August 2026. Until this date, this page described the bulk MOT records as "anonymised" and several research pages attributed them to the Open Government Licence v3.0. Neither was accurate for that DVSA product — the records carry registration marks, and DVSA has published no licence for it — so both statements have been withdrawn. The correction is to wording only: no figure on any page was recomputed or changed by it.

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. How we estimate lifespan and miles remaining

The Estimated Lifespan panel on a vehicle report answers “how much life is left in this car?” from the MOT record itself. It is a survival analysis over 14,837,990 vehicles first used 2003 or later, computed by a committed script rather than typed by hand, and it uses the same definition of “on the road” as our car survival research.

• Off the road means the vehicle’s most recent MOT test is more than thirteen months old. An MOT certificate lasts twelve months, so anything legally in use has been tested inside that window; the extra month allows for late tests.
• Cars still being tested are not thrown away. They count as still going at their current odometer reading — a right-censored observation in an actuarial life table. This matters: taking the average final reading of the examples that have already died would describe a model’s earliest failures, not the model, and would understate a popular modern car by tens of thousands of miles.
• “Miles remaining” is conditional. We do not subtract your mileage from a model average — a car that has already reached 120,000 miles is not an average example any more. We report where half of the examples still going at your reading end up, which is why the figure never goes negative.
• Sample floors, and a named fallback. A model is only scored on its own figures with at least 1,000 vehicles in the record; below that we fall back to the manufacturer’s figures and say so on the page. With neither, the panel does not appear. We publish the sample size on every report.
• We stop where the data stops. Where a model is too young for a level of the survival curve to be observed, we print nothing rather than extrapolate — so a car past everything we can measure is told exactly that instead of being given a number.

Two limits worth stating plainly. The reading we hold is the one recorded at a car’s last MOT, and a car can legally cover another year afterwards, so these are floors on lifetime mileage rather than scrappage figures. And a vehicle exported or declared SORN leaves the MOT record exactly as a scrapped one does, which depresses the figures for models that are exported in volume.

4.1 Why this reads lower than our 16.5-year survival research

Our research post How Long Do UK Cars Actually Last? reports that the median UK car reaches 16.5 years. The Estimated Lifespan panel works from a fleet whose median is 15.8 years. Both come from the same DVSA snapshot (2026-04-12) and the same definition of “on the road”. They differ because they are two different estimators over two different populations, and this is the one place we reconcile them.

The two curves do not cross or contradict; they sit about a year and a half apart. At 16.5 years — the age the research post calls the median — the life table behind the panel still has 45% of the 2003-and-later fleet on the road. Three things pull the panel’s figure lower, all in the same direction:

• Dating. A car last tested in 2015 and never again is, on the research page, part of a cohort scored at its 2026 age. In the life table it is a retirement at the age it reached in 2015. The same car, the same record, two different ages — and the life table’s is always the younger one.
• Different fleets. The research page includes registrations back to 2003. The panel starts at 2003, because DVSA digitisation only begins around 2005: a 1998 car scrapped in 2002 never entered the digital record at all, so the surviving 1990s registrations are the ones that lasted, and any survival figure computed over them is biased upwards. Cutting at 2003 removes that bias at the cost of a shorter observable history.
• An odometer is required. The panel only scores vehicles carrying a plausible mileage reading (1,000 to 400,000 miles), because the mileage and age ladders are built from one pass over the same rows. Motorcycles are also excluded from the make-level and fleet-level roll-ups. The research page has neither restriction.

Which one should you quote? For “how long does a British car last?” the research figure is the right answer, because it uses the whole fleet and the largest sample. For “how much life is left in this car?” the panel is the right answer, because it is scoped to a make or model, corrected for left truncation, and conditioned on what the car in front of you has already survived. Neither is a revision of the other, and we do not quietly move one to match the other.

5. 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.

6. 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.

7. Changelog

• v1.0 — May 2026. Initial public methodology. Establishes data sources, reliability score recipe, common-faults scoring, exclusions and cadence.
• August 2026 — section 4 added. Documents the lifespan / miles-remaining survival analysis behind the Estimated Lifespan panel. No change to any existing input, weighting or exclusion, so the methodology version is unchanged.
• 26 August 2026 — section 4.1 added. Reconciles the two lifespan figures we publish: the 16.5-year cross-sectional median in our survival research and the 15.8-year life-table median behind the Estimated Lifespan panel. Documentation only: no figure on either page was recomputed or changed, and the methodology version is unchanged.
• 28 August 2026 — provenance corrected, section 1.1 added. Section 1 previously described the bulk MOT records behind our research posts as "anonymised". They are not: they are per-vehicle DVSA MOT History records that carry registration marks, held internally and aggregated before publication. Separately, four research posts and their CSV downloads attributed those records to the Open Government Licence v3.0. DVSA publishes no data licence for that product, and the OGL expressly does not cover personal data, so that attribution was unsupported and has been removed from those posts, their CSV downloads and this page. Section 1.1 now sets out the licence position for each source. Wording only: no input, weighting, exclusion or published figure changed, and the methodology version is unchanged.

Frequently asked questions