Methodology: MOT fail rates at 10 years old

What is measured

• Data. DVSA anonymised MOT test data, test years 2022-2025. Class 4 vehicles (cars, small and medium vans, motorhomes), normal tests (test_type = 'NT') only. Retests are not included. The data covers Great Britain (England, Scotland and Wales); Northern Ireland's MOTs are run by the DVA and are not included. Licensed under the Open Government Licence v3.0.
• Fail. A test counts as failed when its result is F (fail) or PRS (pass after rectification: the car failed, and the defect was fixed at the test station during the test). This is DVSA's own "initial fail" definition. If PRS is counted as a pass instead, all cars at 10 fail 24.1% rather than 30.0%; the sensitivity table shows the effect on the rankings.
• At 10 years old. Age = test year minus first-registration (first-use) year, so "at 10" means the MOT in the calendar year the car turned 10. Cars tested at 10 in 2022-2025 were first registered in 2012-2015. Every result describes cars first registered 2012-2015, not the model on sale today.
• One test per car. Only each car's first normal MOT in that calendar year counts. A car is 10 in exactly one calendar year, so the count is the number of cars tested at 10. 4.4% of all normal car tests at 10 (8,244,642 tests) were a second or later MOT in the same year; these are mostly six-monthly private-hire and taxi MOTs. On all tests the fail rate would be 29.7%.
• Unit. Each row is a model family from the VetTheCar normaliser (mot/model_map.parquet), split into generations where the page build splits it, plus two story-specific splits: Volkswagen Golf at first registration 2013 (Golf VI / Golf VII) and Renault Twingo by engine size (897-999 cc = Twingo III, launched 2014; larger engines = Twingo II). Other families mix generations; the generation notes in the tables say which.
• MOT fails, not breakdowns. An MOT records roadworthiness defects on the test day (brakes, suspension, tyres, lights, corrosion, emissions). It does not record breakdowns, engine or gearbox faults, or repair costs. Copy says "fewest MOT fails" or "lower MOT fail rate than average at the same age".

Denominators (one sentence each)

• Baseline, cars: 7,880,798 cars had their first MOT at 10 in 2022-2025, and 2,360,895 failed: 30.0%. Every "vs all cars" figure uses this baseline.
• Mileage: 7,849,808 of those cars (99.6%) have a valid odometer reading (1-500,000 miles). Cars without one are matched on test year only.
• All class 4 tests at 10: 9,049,507 normal tests, by class: car 8,091,055; van 573,860; excluded 167,192; sports 153,587; unmapped 37,201; unclassified 26,612. ("unclassified" = families without a segment; "excluded" = pickups, large-van bases used as motorhomes/minibuses, imports, taxis.)
• Vans: 547,399 small and medium vans (class 4) had their first MOT at 10; 40.1% failed.

Inclusion

• Cars. A family is a car when the reviewed segment map gives it a segment other than van, pickup, other or mixed. other covers large-van bases whose class 4 tests are mostly motorhomes and minibuses (Sprinter, Ducato, Transit-based, Relay, Master and similar), purpose-built taxis and grey imports.
• Ranked: car families with at least 5,000 cars tested at 10: 170 models.
• Sports cars (segment sports: 911, F-Type, MX-5, Z4, SLK, TT, Boxster and others) are ranked in their own table (8 models). They are rarely driven and often garaged, so they are not in the main top 10. They still count in the all-car baseline.
• Vans (segment van) are ranked separately against vans (18 models). The Berlingo, Partner, Doblo, Kangoo and Caddy families include their passenger (people-carrier) versions, which DVSA data cannot separate; the van table marks them.
• Story-specific clean-up of the raw model names (counts are tests at 10):
• MINI: 5 tests filed under the MINI hatch family whose raw model name names another body (e.g. "COOPER COUNTRYMAN") were moved to that family (Countryman 5). The MINI row is "MINI hatch and convertible".
• Camper conversions: Volkswagen California 2,833, Fiat Ducato 98, Ford Transit 27, Peugeot Boxer 26, Mazda Bongo 14, Citroen Relay 14, Mercedes-Benz Sprinter 12, Volkswagen Transporter 9, Volkswagen Crafter 7, Renault Master 6, Vauxhall Vivaro 5, Vauxhall Movano 3, Iveco Daily 2, Peugeot Expert 1, Renault Trafic 1, Peugeot Traveller 1 (3,059 tests at 10 in total) had a raw model name containing CAMPER, CARAVAN, MOTORHOME, CALIFORNIA, KARMANN, WESTFALIA or similar and were removed from their family. Campers registered under a plain name (e.g. "TRANSPORTER") cannot be identified and remain.

The comparison: same test year, same mileage (indirect standardisation)

Fail rates differ by test year (all cars at 10: 32.1% in 2022, 31.3% in 2023, 29.6% in 2024, 27.5% in 2025) and climb steeply with mileage. Because each first-registration year reaches 10 in exactly one test year, a model's mix of registration years sets its mix of test years. Every model is therefore compared with all cars tested in the same year at the same mileage:

expected fails = Σ over the model's cars of the all-car fail rate for that car's test year and 5,000-mile band (61 bands, 0-500,000 miles; missing mileage = its own band)

ratio = observed fails / expected fails · adjusted rate = ratio × 30.0%

The adjusted rate reads as "the model's fail rate if it had the average car's mileage and test-year mix". It is compared only with the baseline, not used to imply a measured share of tests. The raw rate is always shown next to it.

Mileage mix and fail rates of all cars at 10 (for reference):

Why this method and not v1's direct standardisation: direct standardisation projects each model onto the national band weights. For low-mileage models (sports cars, city cars) that means extrapolating from a handful of high-mileage cars, and v1 had to merge sparse bands. Indirect standardisation uses only the bands where the model actually has cars, so it does not extrapolate. The direct method remains as a sensitivity row.

Uncertainty

• Raw rate: Wilson 95% interval.
• Ratio: Wilson 95% interval of observed / cars tested, scaled by cars tested / expected. The baseline (several million tests per test year) is treated as fixed.
• Rank range: 4,000-draw parametric bootstrap (seed 20261007): the 2.5th to 97.5th percentile of each model's rank. A model is "the fewest" or "the most" only if no other model's range reaches that place; otherwise the tied group is named.
• Pairs ("the Corsa fails x times as often as the Jazz"): log-ratio delta method from both models' intervals.
• The intervals cover sampling error only. They do not cover who owns the car, where it is driven, how it is maintained, or the method choices in the sensitivity table, some of which move individual results by several points. Neighbouring models in the middle of the table should not be ranked against each other.

Verdicts

Ages 9 and 11 use the same method with their own baselines (26.6% and 33.1%). They are mostly the same cars a year earlier and later, so this is a consistency check, not an independent replication.

Verdict counts: about average: 15; fewer (<20%): 41; fewer (>=20%): 42; more (<20%): 30; more (>=20%): 32; unconfirmed (< 2,000 cars tested at 9 or 11): 10.

Sensitivity: what changes under other reasonable methods

In the top 10 under every adjusted method: Porsche Cayenne, Audi Q7, Audi A7, BMW X5. In the bottom 10 under every adjusted method: Renault Megane, Dacia Sandero, Citroen DS3, Renault Clio, Chevrolet Spark. We name only results that hold under the headline method and the test-year-matched alternatives, and say where a result depends on whether PRS counts as a fail.

Explanations we disclose but do not adjust for

• Fuel. At 10, cars fail by fuel (raw): hybrid 21.4%, other 26.6%, petrol 29.8%, diesel 30.5%, electric 32.6%. The Lexus models near the top (RX, IS, CT) and the Prius are hybrids; hybrids as a group fail less, probably partly through less brake wear (regenerative braking). The "same fuel" sensitivity row compares each model only with cars of the same fuel.
• Private hire. Models where at least 6.25% of normal MOTs at 10 were a second MOT in the same year (all cars: 4.4%; the share is unrelated to the fail rate across models, correlation 0.01) are flagged (†) as widely used for private hire: Toyota Prius 36%, Mercedes-Benz E-Class 6%, Toyota Avensis 7%, Toyota Auris 7%, Volkswagen Sharan 9%, SEAT Alhambra 7%, Ford Galaxy 7%. Their repeat MOTs are removed, but the cars, and their high usage, remain.
• Upkeep. Owners of premium cars are more likely to have them serviced, often by a dealer, before the MOT, and repairs made just before a test leave no trace in MOT data. The data cannot separate design from upkeep.
• Survivorship. Only cars still on the road at 10 appear. Cars scrapped earlier are missing for every model.
• Region. Corrosion and suspension fails vary by region (road salt, coast). The "postcode area" sensitivity row also matches on the area where the test took place.

What they fail on (components-age10.csv)

For each model, the share of cars failing on at least one item in each of the 11 top sections of the DVSA inspection manual (note: DVSA's "body, chassis, structure" section also holds drive-shaft CV-joint boots and exhaust mountings, so it is not a corrosion measure), compared (observed / expected) with all cars at the same test year and mileage. A component excess is quoted only when the ratio CI clears 1.2 (or 0.8), at least 100 cars had the fail, at least 100 were expected, and the page build's (unadjusted) rate is on the same side of average at ages 9 and 11.

The "non-maintenance" sensitivity row counts only tests with result F that have at least one fail item outside tyres, lights and electrics, visibility (wipers, washers, mirrors) and number plates, so owner-neglect items and PRS fixes do not drive the ranking.

Limitations

• Generations. Families pool 2012-2015 cars and some span a generation change (Clio III/IV, Corsa D/E, Qashqai J10/J11, MINI R56/F56, S-Class W221/W222, X5 E70/F15, Range Rover L322/L405, 911 997/991). The Golf and Twingo are split. Generation notes come from manufacturer launch dates and were checked against manufacturer, press and Wikipedia sources.
• Class 4 only. Most large vans (Transit, Sprinter, Crafter, Master) are tested as class 7 and are not in the van table. The van findings describe small and medium vans.
• Mileage is the odometer reading recorded at the test. Clocking and typing errors exist; readings outside 1-500,000 miles count as missing.
• Badges. The normaliser files the DS3, DS4 and DS5 under DS. 2012-2015 cars mostly wore Citroen badges (the DS 3 badge arrived with the facelift announced in February 2016), so they are named "Citroen DS3" and counted under Citroen in the brand table.

Licence and attribution

Contains public sector information licensed under the Open Government Licence v3.0 (DVSA anonymised MOT test data). Analysis © VetTheCar. Reuse the CSVs and charts with the credit "Source: VetTheCar analysis of DVSA MOT data (OGL v3)" and a link to https://vetthecar.co.uk/best/fewest-mot-fails-at-10-years