Methodology

1. Source data #

We use the Driver and Vehicle Standards Agency's Anonymised MOT tests and results dataset (2023–2025), published under the Open Government Licence v3.0. It contains one row per completed MOT test — vehicle make, model, first-use date, odometer reading, result — plus a separate file listing every recorded defect (RFR) against each test.

Current build: 111,079,869 tests, 12,797 vehicle-year pages, 2,351 model pages and 117 make pages (15,265 URLs in total). Snapshot generated 2026-11-05.

2. What counts as a test #

• Class 4 only (cars, light vans and taxis up to 3,000 kg) — not motorcycles, HGVs or buses.
• Completed tests only: passes (P), passes with rectification at the station (PRS) and failures (F). Abandoned, aborted and refused tests are excluded.
• The yearly extracts overlap, so records are de-duplicated on the DVSA test ID before anything is counted.
• Pass rate = passes ÷ all completed tests for that vehicle year, including retests.
• First-attempt pass rate = passes on normal (non-retest) tests only. This is always lower.

3. Vehicle years and sample size #

A "model year" is the year the vehicle was first used, taken from the DVSA first-use date — not the calendar year of the test. Years 2000–2026 are included. A page is only published when at least 50 tests back it; below that, one unlucky car can move the number several points.

Make and model strings arrive inconsistently from the test stations ("VW" vs "VOLKSWAGEN", "MERCEDES BENZ" vs "MERCEDES-BENZ"). We normalise variants to one manufacturer, ignore placeholder values such as "UNKNOWN" and "OTHER", and slug the rest. Display names are ours; the underlying values are untouched.

4. Failure items #

Each defect is a reason-for-rejection (RFR) code recorded against a test. We join the code to the DVSA item descriptions, then report the share of tests on that vehicle that recorded the item.

Severity comes from the tester's own recording: Dangerous and Major mean an immediate fail; Minor and Advisory do not. A test can carry several items, so percentages do not sum to 100%. Failures and advisories are ranked separately, because advisories are far more common and would otherwise bury the items that actually fail cars.

The Est. cost shown next to a failure item is a typical UK part or repair price range — an editorial estimate refreshed with each build. It is not derived from the DVSA dataset, which contains no pricing, so treat it as a rough budget for the part, not a quote.

5. Mileage #

Mileage comes from the odometer reading at test. We discard readings below 100 miles (usually mis-keyed) and above 350,000, then group tests into 0–30k, 30–60k, 60–90k, 90–120k and 120k+ buckets. "Typical mileage" on a page is the median of those readings, which is more robust than the mean.

6. Fuel, regions and seasons #

Fuel type is the DVSA code recorded at test time, grouped for display into petrol, diesel, electric, hybrid, LPG and other. Electric and hybrid volumes are still small in the years we cover, so their pass rates come from low test counts on many pages.

Regions are grouped from the postcode area of the test station (e.g. SW, B, BT). A few areas straddle a boundary; each is assigned to the region holding most of its population. Only regions with at least 15 tests on a vehicle appear on its page, so a single odd garage cannot skew the picture.

Winter vs summer uses the calendar month the test was recorded (October–March vs April–September). It is reported only when a vehicle has at least 30 tests on each side. We show the test month, not the vehicle's expiry month, because expiry is not in the public data.

7. Repair cost estimates #

The DVSA data records what failed a car, never what the repair cost. The money figures on vehicle pages are our estimate, built in three steps:

1. Every failure item is mapped to the part that has to be replaced or adjusted — a coil spring fractured maps to a coil spring, a stop lamp not working to a bulb. Items that cannot be mapped to a part are left out rather than guessed.
2. Each part carries a price bracket for typical UK supply, and for fitting where fitting is not a DIY job. Bulbs are £3–£12, a coil spring £25–£65, welding for corrosion £180–£450.
3. The brackets are averaged across the failure items recorded on that model, weighted by how often each item appears. A model whose failures are mostly bulbs gets a low average even if it fails often; a model whose failures cluster in suspension and brakes gets a high one.

What this deliberately leaves out: diagnostics, a retest fee (currently up to £54.85, but many garages include it), garage labour rates beyond the fitted estimates, and anything the tester did not record. Prices vary by region and by whether you use a dealer, an independent or your own tools. Treat the figure as a shortlist of what to budget, not as a quote.

These brackets are editorial estimates maintained alongside the pipeline, not figures from the DVSA dataset, and they are excluded from the pass-rate calculations entirely.

8. What people actually paid #

Estimates are only as good as their source. The calculator asks one optional question — what did you actually pay? — and the answers replace the estimate on the price list once there are enough of them.

• What is stored: a failure item, a whole number of pounds, and a timestamp. No name, no email, no registration, no vehicle, no IP address, no cookie.
• What is rejected: anything that is not a whole number from £1 to £10,000, and any item that is not one of the published failure pages.
• What is published: only the median per item, and only after 10 reports. Medians rather than averages so one engine rebuild cannot drag a figure, and a threshold so a single anecdote never appears as a price.
• What is not claimed: that these are quotes, or that a report has been verified against an invoice. They are self-reported, and labelled as such.

Unlike every other table on this site, this one cannot be rebuilt from the raw DVSA extracts, so re-running the pipeline replaces the DVSA data and leaves the reported prices alone.

9. Comparing like with like #

A raw pass rate is not a fair comparison. A 2006 saloon with 150,000 miles and a 2021 city car with 20,000 are not the same exam, so every vehicle page shows the figure it should have — and the difference from it.

Two things move a pass rate more than the badge on the bonnet: age and mileage. Both are modelled.

1. Age. Each page is compared with the national pass rate for its own year of manufacture. Cars get harder to pass as they age, so comparing a 2004 car with a national average dominated by newer cars would be meaningless.
2. Mileage. The national pass rate is calculated for every combination of year of manufacture and mileage band (0–30k, 30–60k, 60–90k, 90–120k, 120k+, from the recorded odometer reading at the test). The expected rate for a vehicle is the weighted average of those bands, using how many of its tests fall in each one.

That method is called indirect standardisation. Using each vehicle's own mileage mix matters: a van that spends its life at 120k+ should not be marked down for being a high-mileage van, only for failing more often than other vehicles of the same age and mileage. Across the site the adjustment moves a verdict by about one point on average and up to around three and a half, so it changes close calls rather than rewriting the table.

Because the adjustment is applied to every model, it can also be used to rank them — the models that beat their own age benchmark, and the ones that fall furthest behind it.

Fuel type and how the car is driven are not adjusted for, because the data does not allow the two to be separated cleanly at model level. The residual is therefore "better or worse than comparable cars", not "well built".

10. Postcode areas #

Every MOT test in the DVSA extract carries the keeper's postcode area — the letters at the start of a postcode, such as KY, E or CF. There are 118 areas with enough tests to publish. Each one gets a page comparing its pass rate with the national rate for the same mix of model years and mileage — and the same model year is compared across areas as a second, blunter control on the same question.

The adjustment here is direct standardisation, the mirror of the method above: every area is given the national distribution of model year × mileage band, and its own pass rate within each cell is applied to those weights. What survives is not "old cars" but how cars in that area are kept and tested. Doing it this way makes the gap wider (about 16 points between the highest and lowest area) than the raw gap (about 14), which is the opposite of what an age artefact would do.

Three things this is not:

• Not a garage ranking. The anonymised results have no test-station field, so we can list stations (from the DVSA's published active-station list — see the station directory) but not measure them.
• Not where the test happened. It is where the keeper is registered. A fleet car registered in one area can be tested anywhere.
• Not a town. A postcode area covers many towns; the label on each page is the town with the most MOT stations in it, which is a convenience, not a definition.

Per-area model tables use a 1,000-test floor and only include models that clear the site's 2,000-test floor nationally, so a single unlucky garage cannot invent a local worst-seller.

11. Known limitations #

• This is a historical snapshot, not a live lookup. It cannot tell you about one specific car.
• MOT rules changed in May 2018 (defect categories and the exemption for cars under three years old), so item-level comparisons across that line are imprecise.
• Test outcomes reflect owners and garages as much as engineering: a well-maintained average car beats a neglected "reliable" one.
• Make and model names are self-reported at test time; rare imports and grey imports are inconsistent.
• Pass rates for very young vehicles reflect low test volumes and the three-year exemption.

13. The email list #

The Keep me posted form on model, failure and calculator pages, the footer box and the email updates page all feed one list. We store the address, what you signed up about, the page you were on and the date, and only after you tick the consent box. The list lives in our own database and is exported by us — nothing is sold or shared with a partner.

It is a double opt-in. The address is not on any list we would mail until you click the link in the single confirmation email that follows. Until then the row sits unconfirmed. Every later email carries a one-click unsubscribe that works without logging in, and re-submitting the form never re-subscribes an address that unsubscribed. One address can follow several things (a few models, a failure, the site as a whole); those are separate rows, and each one is removed independently. To be removed by hand, write to hello@motfaults.co.uk.

14. Corrections #

If you think a figure is wrong, email hello@motfaults.co.uk with the page and the number you expected. We publish the method, not the marketing.