UKCheck a car
Menu

Methodology

How we turn MOT tests into the figures on this site

DVSA MOT tests 2022–2025 · 130,233,413 MOT tests · Updated

Every MOT figure on VetTheCar comes from 130,233,413 DVSA MOT tests (2022 to 2025), and every comparison is with cars tested at the same age [1].

  • 1,000 MOT tests per model and age before we show a model's fail rate; advisory follow-up figures need 200 pairs of tests.
  • 95% intervals: a page says a model or a part does better or worse than cars the same age only when the interval supports it; otherwise it says "about the same" or makes no comparison.
  • One date: every page shows the day the site's data was last updated, 7 Oct 2026.

The data we use

DVSA publishes the result of every MOT test in Great Britain, without registrations or keeper details. We use the class 4 normal tests from 2022 to 2025 of vehicles aged 3 to 25 at the test: 130,233,413 MOT tests [1]. Class 4 covers cars and light vans up to 3,000 kg.

Recall pages use DVSA's vehicle recall records [2]. Repair cost estimates use the Office for National Statistics' local price quotes for car repairs [3]. Every data source we use is listed with its licence.

The data is licensed under the Open Government Licence v3.0. It lets anyone use it, as long as the source is credited and the use does not suggest that DVSA endorses the result. We are not affiliated with DVSA, DVLA or any car maker.

Which tests count

  • Normal tests only (DVSA test type NT); partial retests (RT) are left out. A car that comes back for a full new test counts again.
  • A test counts as a fail when DVSA recorded a fail, including a fail fixed at the test station within an hour (PRS).
  • Age is the test year minus the year the car was first used. Tests without a first-use date are dropped. We report ages 3 to 25.
  • Typical mileage is the median of the mileage recorded at those tests.
  • Fail items are counted per 100 MOT tests of that model at that age. A test with two faults counts twice, so this is not a share of cars.
  • What happens after an advisory comes from 83,372,892 pairs of consecutive normal tests of the same vehicle, 300 to 500 days apart (about 10 to 16 months). A pair counts when the first test advised on a part, or gave it a minor defect, and the second failed on the same part. A figure needs at least 200 pairs.

From typed model names to model families

Testers type the model name, so one model appears under many spellings and trim names. Fixed rules group them into 780 model families: "FIESTA ZETEC" is a Fiesta and "FOCUS TITANIUM" a Focus. Where a maker reused a name for a different car, such as the Ford Puma of 1997 and of 2019, the family is split by first-use year, which gives 803 family and generation groups.

129,340,595 of the 130,233,413 MOT tests (99.3%) match a model family. The rest still count in the figures for all vehicles.

We name an engine or a generation only where a reviewed source row says which one it is. Until then a page uses the plain label, such as "1.0 petrol (999 cc)" or "first used 2014 to 2017".

What every figure is compared with

A model is only ever compared with cars tested at the same age. A 10-year-old car fails more often than a 4-year-old one, so a comparison across ages would say nothing about the model. The baseline is all class 4 vehicles tested at that age [1], until a baseline without vans is built.

Today the MOT fail rate pages set a model's raw fail rate against all class 4 vehicles tested at the same age over the four years. A comparison at the same mileage and in the same test year (a 2013 car seen in 2023 set against all cars tested at 10 years in 2023) is being added to those pages.

All class 4 vehicles by age at test, 2022 to 2025
AgeFail rateMOT testsMedian mileage
3 years12.0%8,016,46322,806
4 years13.4%8,699,93729,800
5 years15.3%9,425,36037,103
6 years17.8%10,382,72044,951
7 years20.6%10,537,31852,953
8 years23.8%10,440,45760,745
9 years27.1%9,949,01268,033
10 years30.3%9,049,50775,043
11 years33.3%8,087,51481,778
12 years36.0%7,303,15187,673
13 years38.3%6,612,16092,305
14 years40.3%6,054,52696,840
15 years41.6%5,591,675100,157
16 years42.6%4,858,625102,757
17 years43.1%4,040,458105,023
18 years43.3%3,261,622105,218
19 years *43.2%2,479,858104,356
20 years *42.8%1,845,952103,393
21 years *42.0%1,321,417101,873
22 years *41.0%913,980100,352
23 years *39.6%619,53198,905
24 years *37.8%428,81597,398
25 years *36.3%313,35597,082

* Only survivors are still tested at this age: cars scrapped earlier are not in the data, so these rows describe the cars still on the road. We give no verdict beyond 18 years.

Mileage standardisation

A model that is driven further than average fails more often for that reason alone. Where a page says a rate is "adjusted for mileage", we split the model's MOT tests at that age into six mileage bands (under 30,000; 30,000 to 60,000; 60,000 to 90,000; 90,000 to 120,000; 120,000 to 150,000; over 150,000 miles) and weight each band by its share among all class 4 vehicles tested at the same age. This is direct standardisation. The comparison is all class 4 vehicles the same age in the same bands.

Today the mileage-adjusted rate is used for the popular models on the tools page and the model figures in the sample report. The MOT fail rate pages still show the raw rate and compare it with all cars the same age; the adjusted rate is being added to them.

Rankings order models by their fail rate against cars the same age; the rankings about high-mileage cars compare with cars the same age at the same mileage. Moving every ranking to mileage-adjusted rates is in progress.

A raw rate is the share of all MOT tests of a model at one age that ended in a fail, with no adjustment. Where a page shows one, it is labelled as a raw rate or set only against cars the same age.

Confidence intervals, and when we say "more often"

Raw rates use a 95% Wilson interval: the range the true rate is likely to sit in, given the number of MOT tests. Rates adjusted for mileage use a 95% normal-approximation interval over the weighted bands. A rate from 200,000 MOT tests has a narrow interval; one from 2,000 a wide one.

A page says a part fails "more often" or "less often" than on cars the same age only when all of these hold:

  • at least 100 MOT tests with that part failing, and at least 100 expected at the all-car rate;
  • the 95% interval of the ratio clears 1.2 for "more often" or stays under 0.8 for "less often";
  • the same direction holds at 3 or more ages.

Otherwise the page says "about average". A mileage-adjusted model rate is called higher or lower than cars the same age only when its interval excludes their rate, the gap is at least 2 percentage points and at least 8% in relative terms; otherwise it is "about the same". Ranks such as "198 of 309" are counted among all models with 1,000 or more MOT tests at that age.

When a page is published, and when it is indexed

Two separate tests decide what you can read here. A page is published when its data is honest: enough MOT tests that every figure on it means something. It is indexed by search engines only when the data also has a story: a difference the intervals support. A published page that fails the second test stays readable and linked from its parent, but it is marked noindex and left out of the sitemaps.

Thresholds, from the settings every page is built with
PagePublished withIndexed with
MOT fail rate of a model3 ages with 1,000 or more MOT tests each; an age with fewer is not shown5,000 MOT tests at its most-tested age, and a fail rate at that age whose 95% interval excludes the rate of all class 4 vehicles the same age (today the raw rate; a mileage-adjusted comparison is being added)
Common problems of a modelAs the MOT fail rate pageAs the MOT fail rate page, plus at least 3 parts that pass the comparison rule
One part of a modelMOT tests at its most-tested age, with that part failing at least once2,000 MOT tests at that age and a ratio that passes the comparison rule
One first-use year of a model1,000 MOT tests for the year1,000 MOT tests, 2 ages with 2,000 or more each, and a difference from the years either side that the 95% interval supports
An MOT defect1,000 fails nationally5,000 fails nationally and 500 advisory pairs to measure what happens next
Recalls of a modelOne recall matched to it3 recalls matched exactly, or through our model-name rules (the same ones as the MOT data)
MOT pass rates of an area1,000 MOT tests20,000 MOT tests from 2022 to 2025
Hubs, tools and these pagesWhen they have contentWhenever they are published

Repair cost estimates

Most repair costs are estimates, and every one is labelled "Estimate" with its parts: the labour time for the job, multiplied by an hourly rate from the Office for National Statistics' local price quotes for car repairs, plus parts. Where ONS publishes a price range for the job itself, we show that range instead. [3]

Cost figures never appear in page titles or descriptions, in verdicts, in how serious a fault is called, or in any ranking.

How often the data is updated

DVSA publishes the anonymised MOT data once a year, for the year before. We rebuild the statistics when a new year is published, and whenever a fix to our method needs it. The site's data was last updated on 7 Oct 2026, from MOT tests up to the end of 2025.

Every page shows that one date: the "Updated" line under the heading, the date in its structured data and the date in our sitemaps are the same day. Recall records and clean air zone rules have their own source dates, shown where they are used.

What the figures cannot tell you

  • They are about a model, not one car. A model's fail rate says nothing certain about the car in front of you. Its own MOT history does, and a mechanic's inspection does more.
  • They are not a measure of build quality. Mileage, how a car is used and how well it is looked after all change the fail rate.
  • The MOT does not test everything. Clutches, gearboxes, timing belts and chains, coolant systems and most engine faults are not in the data.
  • Wear items are common. Tyres, brake pads, wiper blades and bulbs are among the most common fails. They say more about upkeep than about the design, so pages show them apart from structural and safety items.
  • Survivors only. Cars scrapped before a test are not in the data.
  • Names are typed by hand. The grouping rules catch most spellings, but some tests end up in the wrong family or in none.
  • Class 4 includes light vans. Van versions of a car are in the same family when testers used the same model name.
  • Recorded mileage can be wrong. A reading typed in kilometres or with a digit missing distorts a median slightly. One car's readings are checked on their own.

Corrections policy

If a figure or a sentence on this site is wrong, we want to know. Tell us the page and what you expected.

  • Email support@vetthecar.co.uk, or use the contact page.
  • We check every report against the source data and reply.
  • When a figure was wrong, we fix the cause, rebuild every page it appears on and record the change below with its date. We do not edit a figure quietly.
  • When a source corrects its own data, we rebuild from the corrected file and note it here.

Corrections log: no corrections have been published yet.

Who is responsible

The figures are built and checked by the VetTheCar data team, a small independent team that also runs a used-car check in Sweden. Tom Ashby, editor, is responsible for this method and for the corrections log.

More about us is on the about VetTheCar page, and the company details are in the footer.

Buying Report £14.99 incl. VAT · one car · no subscription · Not a mechanical inspection.

Sources

  1. Driver and Vehicle Standards Agency, Anonymised MOT tests and results, 2022 to 2025. DVSA MOT data, Open Government Licence v3.0, 130,233,413 MOT tests. Accessed 7 Oct 2026.
  2. Driver and Vehicle Standards Agency, Vehicle safety recalls data. DVSA recall data, file dated 22 Sep 2026.
  3. Office for National Statistics, Consumer price inflation, local price quotes (car repair items). ONS price quotes, September 2025 to August 2026.