How we work

Where it comes from. Each part of the car has a risk index from 0 to 100, calculated from its three most serious faults (severity × frequency, weighted 1, 0.5 and 0.25). The score is 10 − index / 10: an index of 30 equals a 7.0. So a single serious, frequent fault lowers the score even if everything else is fine, and ten minor faults do not sink a reliable car.

Scores by part of the car. Engine, gearbox, suspension, brakes, steering, tyres and wheels, electrical system, electronics and multimedia, air conditioning, bodywork and interior, and airbags and safety. The engine score is the average of the engines fitted to that car; we also show the best and the worst, because choosing the right engine changes everything. The gearbox score includes the documented automatic gearboxes (DSG, EAT, EDC…) and the gearbox and clutch faults of the car itself. If we know which gearboxes the car offered, each variant is scored separately: the manual gearbox with the faults of its gearbox and clutch, and each automatic gearbox with its own. The gearbox score, and that of the rest of the car, is the average of the variants, just as the engine score is the average of its engines: the faults of an automatic gearbox do not lower the score of someone who buys the manual version, nor the reverse.

Overall score. Half comes from the engine and the other half from the rest of the car, calculated from the most serious faults across all its parts. It is not the average of the partial scores: a serious brake defect is not offset by good air conditioning.

No data, no score. A part we have not yet analysed appears as “Not analysed”, never as a 10. If a car has only had its engine analysed, its score is provisional and is marked as such. “No documented faults” means we have analysed it and found no reliable sources with known faults, not that it cannot fail.

Which cars make the rankings. Only cars with solid evidence: at least three sources on their page and five documented faults across the engine and the rest of the car. A car with few documented faults may simply be a little-studied car, not a reliable one, so its score is shown on its page with a “limited evidence” warning but does not compete in the rankings. A car with no studied engine does not compete either: its score would only measure the rest of the car and could not be compared with the others. Nor does one that has been on sale for more than 1 year and whose engines we have not yet studied: its engine score would only measure part of its versions. And a score of 7.0 or higher, which says a car is reliable, requires strong evidence: at least 6 sources in 3 different languages. MOT inspection data alone is not enough, because it does not measure the engine. As with engines, the car must also have been on sale for at least 4 years: a newly launched model has not yet had time for its faults to appear.

Electric and hybrid. In an electric car, the “engine” is its whole propulsion system: the traction battery, charging, the electric motor with its reduction gear and the power electronics. It is scored with the same formula. Pure electric cars compete in the general car ranking and also in a ranking of their own, so that they are compared only with each other. Their electric motors have their own engine ranking, because their faults cannot be compared well with those of a combustion engine. Hybrid engines are petrol engines with their own ranking, and they also stay in the petrol engine ranking.

Sorting by part. In the car rankings you can sort by overall score, by engine only or by each part of the car. When sorting by a part, cars with no documented faults in it are listed separately, without a score: not finding any faults is not the same as scoring a 10.

Scale. 7.5 or more: good reliability. 5 to 7.4: acceptable. 2.5 to 4.9: poor. Less than 2.5: very poor. Faults fixed free of charge under a recall carry less weight, because once the campaign has been carried out the defect is resolved.

For the cars we analyse in depth we also record the faults that do not depend on the engine: suspension, brakes, tyres and wheels, steering, gearbox and clutch, electrical system, electronics and multimedia, air conditioning, bodywork and interior, and airbags. They follow the same rules as engine faults: open, cited sources, severity and frequency on the same scales, and costs in Spain including VAT.

Whole-car index. We calculate two indexes from 0 to 100 using the same formula as for engines (the three most serious faults, weighting severity and frequency): one from the average risk of its documented engines, counting only the faults that affect the versions fitted to that car, and another from its non-engine faults. The car’s score is the average of the two; if we have not documented any of its engines, only the second counts.

Minimum evidence. We only score the rest of the car, and it only enters the rankings and comparisons, when we have found at least 2 sources and 2 documented faults outside the engine. With fewer, the lack of faults says more about what we have been able to research than about the car: we show what we have found, but its score stays provisional (engines and gearbox only) until the analysis is complete.

Recalls. A fault that the manufacturer fixes free of charge under a recall or a technical campaign counts in the car’s index with a frequency of “occasional” at most, even if its page lists it as acknowledged by the manufacturer: on a car where the campaign has been done, the defect is fixed. That is why, before buying, you should always use the VIN to check that the campaigns have been carried out.

Rankings by system. In each system (for example, suspension) we rank the analysed cars by the index of their faults in that system. A car having no documented faults in a system does not guarantee it will not fail: it means we have not found reliable sources describing it as a known problem.

Comparisons. We highlight comparisons between rivals of the same segment and era; the other combinations can be viewed, but we do not suggest them to search engines.

The Dutch vehicle registry (RDW) publishes every periodic technical inspection (APK) and every defect the inspector finds as open, public-domain data. For each generation we take a sample of cars by registration year within their years on sale, leaving out the years when it overlaps with another generation, and count from 2022 which parts had any defect in their inspections: suspension and axles, brakes, steering, lights, rust and bodywork, exhaust, leaks, glass and tyres. Advisory notes that are not defects do not count.

Compared with cars of the same age. A twelve-year-old car has more defects than a four-year-old one, so each inspection is compared with the average of all the models in the sample at that same age (vans, which cover far more miles, only against other vans). A part's index is 1 if it has the normal defects for its age, 0.5 if it has half, and 2 if it has double. So that a few cases do not inflate the index of parts that rarely fail, we bring it closer to 1 by adding 10 expected defects to both the observed and the expected counts. We only publish a generation's figure when it has at least 300 inspections.

How it counts in the score. The index is converted to the score scale: the normal level for its age equals 7.5, double the defects equals 5, and triple equals 2.5. For suspension, brakes, steering, lights and bodywork, the part's score is the average of its documented faults and the inspection; if there are no documented faults, the inspection alone is enough to score it. In the index for the rest of the car, the inspection counts for a quarter and documented faults for three quarters.

Limits. The inspection measures wear and defects visible at a check, not engine, gearbox or electronics failures, which still come from the documented sources. These are cars registered in the Netherlands that have survived to their inspection, and the registry does not give mileage: cars with low use (sports cars, convertibles) tend to do better and heavily used ones (pick-ups, taxis) worse. Tyre condition depends mostly on maintenance and does not count in the score.

Which tests count. Only multi-product comparisons carried out by independent testers with their own measurements count: motoring clubs (ADAC, ÖAMTC, TCS, RACE, ANWB), consumer associations (Stiftung Warentest, OCU, Which?, Altroconsumo…) and magazines that run their own tests (Auto Bild, auto motor und sport, Auto Express…). Shop listings, sponsored content, what the manufacturer says and opinions on a single product do not count. A single test published by several organisations (the ADAC test with its European partners) counts only once.

How they are ordered. In each test, the first product scores 10 and the last scores 0, in proportion to its position among all the products tested. A product’s score is the average of its recent tests (the last three years for tyres, because models change every few seasons; more years for the rest). To enter the ranking, a product must appear in at least two tests, and a category is only published with at least three products that meet this condition: with a single test, the ranking would simply repeat that test. Each result shows the position, the original rating from the test and a link to the publication.

Photos and shops. Each product’s photo is taken from its manufacturer’s official website, linked from there and credited; the price listings come from the shops themselves. Shop links are affiliate links: if you buy, we may receive a commission, but the order is decided only by the tests, and no brand pays to appear.

Tyre sizes, battery, wiper blades and oil come from the owner’s manual, from the manufacturer’s catalogues and price lists, or from the fitment databases of spare-parts manufacturers. A tyre pressure is only published if it comes from the car’s manufacturer or from two independent technical sources that agree; otherwise it appears as ‘no reliable data’. The final reference is always the label on your car (door frame or fuel flap), because it changes with the engine, the wheel and the load.

The scores and rankings are the result of applying the method on this page to documented, cited facts: a reasoned assessment of what the sources say, not our own test of each car or a claim about a particular unit. We write in neutral terms, attribute each fact to its source and do not use brand logos, only their names to identify products.

If you are a manufacturer, importer or brand and you believe a figure is inaccurate or out of date (a fault corrected in production, a campaign that does not appear, versions wrongly assigned), send us the documentation. We review it using the same criteria as any other source, normally within 15 days: if the figure was wrong, we correct it and say so; if it stands, we can add your documented position to the page.