Every figure on Howreliable? comes from DVSA's record of every MOT in Great Britain. Each model is compared with cars of the same age and mileage, so a model isn't marked down just for being older or more heavily used.
DVSA's anonymised MOT tests and results for 2025: 31,475,521 first tests on cars (MOT class 4) first used from 2005 to 2022. DVSA publishes the data under the Open Government Licence; we use it as published and don't add data from anywhere else.
Which tests count. Each car's first MOT of the year. Retests after a failure are left out, so a car that fails and is retested counts once.
Passed first time. The test ended in a pass. A car that failed but was fixed within the hour, before the result was recorded, counts as a fail, as DVSA counts it.
Registration year. The year the car was first used, which DVSA records for every test.
Model. DVSA records make and model as text, often with the trim or engine code ("118D M SPORT", "C 220 AMG LINE", "R ROVER SPRT"). We match each model's names with a pattern; every model page says how many names matched, with examples. 101 models are covered so far.
Older cars and cars with more miles fail more often, so comparing a model with all cars would mostly measure its age. Instead we put every car into a group by registration year and mileage (under 30,000, 30,000 to 60,000, 60,000 to 90,000, 90,000 to 120,000, and over 120,000 miles).
A model's comparison figure is what those groups scored, weighted to match the model's own mix of ages and mileages. "−4.3 points" means the model passed first time 4.3 percentage points less often than cars of the same ages and mileages.
Report-card dots compare how often a model's cars of one registration year failed on one area of the test (brakes, suspension and so on) with comparable cars. Much worse: at least 1.5 times as often. Worse: 1.15 to 1.5 times. About the same: 0.87 to 1.15 times, or less than half a percentage point apart. Better: 0.67 to 0.87 times. Much better: two-thirds as often or less.
Year groups sort registration years by their first-time pass rate against comparable cars. Well behind: 5 or more points below. Behind: 1.5 to 5 points below. Level: within 1.5 points. Ahead: 1.5 or more points above. Neighbouring years in the same group are shown together.
Failures and advisories. A failure is a defect that failed the test (including ones fixed within the hour); an advisory is wear the tester noted that didn't fail it. "Fails more often than other cars" lists defects found on at least 0.2% of a model's MOTs, with the ratio to comparable cars.
Rivals. Each model is compared with models of the same type (city cars, family SUVs and so on), using cars registered 2014 to 2021 so every model is measured over the same years. Models with fewer than 500 MOTs from those years aren't ranked.
We don't give cars a score or a star rating: a single number hides which years and which parts are the problem, and we couldn't show you how it was made. We show the percentages and the comparison instead.
We don't use owner surveys, warranty claims or reviews, and we don't write anything about a car that its MOT results don't show.
DVSA anonymised MOT tests and results, open.data.dvsa.gov.uk. Contains public sector information licensed under the Open Government Licence v3.0. Howreliable? is not connected with DVSA or any carmaker.