How we count

Inspection failure rates

• First inspections only. A car that fails, is repaired and comes back would otherwise be counted twice. We use New York's initial inspection records and the result of the safety check.
• Twelve months. City and state rates cover inspections since October 2025; model rates cover August 2021 onwards.
• Like with like. Older cars and cars with more miles fail more often, so we show rates by vehicle age and by mileage next to the overall figure, always against the state average.
• Models and places, not businesses. We publish failure rates for models, cities and the state — never for a single inspection station. A station's rate mostly reflects which cars its customers bring in, and it would read as a verdict on a named business. Station pages show volume and the mix of cars instead.
• A city means the inspections carried out at public inspection stations licensed in that city, not where the car is registered. Example: Albany — 123,331 first inspections, 4.11% failed; New York State 1.46%.
• Minimum size. A model needs at least 150 first inspections in a city before its rate is shown there.

Mileage and odometer checks

• US: the odometer reading recorded at each New York inspection. A record is consistent when no reading is lower than the one before.
• Netherlands: the RDW publishes its own judgement (logisch or onlogisch) rather than the readings; we show that judgement as it is.

Fatal crashes

• US figures come from NHTSA's Fatality Analysis Reporting System, a census of every crash on a public road in which someone died within 30 days. We count each vehicle once per crash, by its make, model and model year as NHTSA decoded them.
• Counts, not rates. FARS records crashes, not how many cars are on the road or how far they are driven, so a popular model appears more often. We do not turn these counts into a safety ranking.
• Deaths are people who died inside that vehicle. Pedestrians, cyclists and people in other vehicles are not added to it.
• Great Britain figures come from the Department for Transport's STATS19 records of collisions reported to the police in which someone was injured. We count each car once per collision under the generic make and model the DfT assigns; collisions where the make and model were withheld are left out.

Real-world fuel consumption

• Every new car sold in the EU since 2021 records the fuel it burns and the distance it covers. Manufacturers and national authorities pass these lifetime totals to the European Environment Agency, which publishes them car by car without any registration or owner details.
• We add up the fuel and the distance of all cars of a model and divide one by the other, so a car that has been driven further counts for more. Only the records the EEA itself uses in its calculations are included, and a model needs at least 100 cars.
• We also leave out cars with fewer than 1,000 km on the meter and readings that cannot be right: a petrol or diesel car using less than half or more than two and a half times its official figure, or a plug-in hybrid above 15 l/100 km. This removes about 1% of the cars — mostly whole batches reported with the wrong figures.
• The official figure is the car's WLTP CO₂ value converted to litres with the factors the EEA's own tables imply (22.78 g/km per l/100 km for petrol, 26.31 for diesel).
• Petrol and diesel cars, hybrids that cannot be plugged in (mild hybrids included) and plug-in hybrids are kept apart. For plug-in hybrids the figure covers all the distance driven, including on electricity, which is why it is so far above the official value when owners rarely charge.
• Vans (light commercial vehicles) report the same way and are shown separately from cars. On model-year pages the year is the year of first registration.
• Fuel cost. The calculator multiplies a model's real-world consumption by the latest weekly average pump price, taxes included: the European Commission's Weekly Oil Bulletin for EU countries (in euros, also outside the euro area), the UK Department for Energy Security and Net Zero, and the US Energy Information Administration. Plug-in hybrids are left out because much of their energy comes from the grid, and charging costs are not covered.

Comparing two models

• A comparison puts the same records for two models side by side. Any two models with a page can be compared; the pairs we list are rivals: the same body type (car, SUV, sports car, minivan, pickup or van), within about 15% of each other's average weight, on sale in at least three of the same years, from different makes, and with at least four of our measures on record for both.
• Body type comes, in this order, from Wikidata's class for the model, Natural Resources Canada's vehicle class, the body codes of New York registrations and the body recorded by the RDW for Dutch cars. Weight is the average mass of new cars registered in Europe (less 250 kg where most are electric, for the battery), or an estimate from the Canadian size class.
• "Where each one leads" only uses measures that do not grow with sales: MOT and New York inspection results against cars of the same age, NHTSA star ratings, official and real-world fuel use, official CO₂ and the share of Dutch cars with an illogical odometer record. Differences smaller than a set margin (for example three points on the MOT measure) are called too close.
• Complaint, recall and service bulletin counts are shown for both models but never declared better or worse: a model that sold twice as many cars collects about twice as many.

Driving test centres

• British test centre figures are DVSA's own: car tests conducted and passed at each centre by financial year (April to March) and month, first attempts, and the median weeks between booking and taking a test.
• Test centres are run by DVSA, a government agency. We show their published pass rates but do not rank centres as "easy" or "hard": a pass rate mostly reflects who tests there.
• Centres with fewer than 100 tests in the latest year have no page; DVSA itself leaves out centres where only one examiner tested.

MOT test stations

• The list of garages approved to carry out MOT tests is DVSA's own and is updated every three months. Telephone numbers in it are not used.
• A station is linked to a business in our directory only when both have the same postcode and share a distinctive word in their names. Stations without a match are listed under their DVSA trading name, in the town where businesses with the same postcode are.
• Approval is not a rating: we do not publish MOT results for individual stations.

Which pages exist

• A page exists only when there is enough data to fill it. A model year has its own page when at least three independent data blocks cover it — for example NHTSA complaints, recalls, service bulletins, New York, Dutch, British, Finnish or New Zealand inspection and registration figures, fatal crashes, real-world fuel use or new registrations in Europe. A model has a page when three such blocks cover the model as a whole or when one of its years has a page; a make has a page when one of its models has. Everything else is listed, without a link, on the page above it.
• In the directory, a town or city has its own page from three businesses. A page for one kind of business in a town (its fuel stations, say) exists only when the town has at least three of them and other kinds of businesses too; where every listed business is of the same kind, the town page already shows them all and the address points there.
• New York and Connecticut places have their own page from three licensed businesses. The DMV registers spell many places in several ways (Long Island City appears as LONGISLANDCITY, LONG ISLND CTY and more than twenty other forms); spellings in the same county are merged when they differ only in spacing, in an abbreviation (PK for Park, PT for Port or Point, whichever exists) or by one letter in a rare spelling, and the place takes its usual name. Smaller places are listed with their businesses on the state page.
• Pages are linked to each other through the data they share: a recall links to the model years it covers and to recalls of the same models in other countries, an inspection defect to the models where it is found most often, a Dutch approved business to the same business in the directory (same postcode and a shared distinctive word in the name), a French driving school to fuel prices, inspection centres, speed cameras and the low-emission zone of its commune.

Matching models across countries

• Make and model names differ between registers. We match them by rules — names normalised, VINs decoded by NHTSA for a sample — and check the result: 97.1% of a 6,000-VIN sample agree with NHTSA's decoding.
• When a match is uncertain we keep the models apart rather than merge them.
• European registers list Mercedes-Benz and BMW cars by trim (C 220 d, 320d). We group European figures under the class or series (C-Class, 3 Series); pages for individual US trims keep their US records.
• Models sold mainly in Europe get a page when at least 5,000 were registered new in the EU, Norway and Iceland between 2010 and 2022 (European Environment Agency). Their names and the list of real models come from Wikidata, so a garbled entry in the registration data never becomes a page of its own. A model year of these cars gets its own page when at least three European sources cover it: new registrations, UK licensing, MOT tests, Finland and New Zealand.

Generations

• Generations and their production years come from Wikidata. We use them only where Wikidata links the generations to one another and gives their start year; a generation without an end year ends when the next one begins.
• A model year belongs to every generation whose production run includes it, so a changeover year can count towards two. A generation gets its own page once at least two of its model years have one.

How common is it

• We count a model's cars licensed in the United Kingdom, registered in the Netherlands and New Zealand and in use in Finland, and rank it against every model on CarsLore in that country. Fewer than 200 cars is "very rare"; otherwise the label follows the rank, from the top 2% ("one of the most common") down.

Engines

• Engine versions come from the Dutch register: engine size rounded to 0.1 litre, net maximum power and fuel of every registered car. A version is shown when at least 20 cars of that model year have it; horsepower is kW × 1.341.

New car market

• The market pages are also written in German, French, Dutch, Italian, Spanish, Portuguese, Polish and Swedish for the countries where those languages are spoken.
• The market pages count new passenger cars registered for the first time, as each country reports them to the European Environment Agency: every EU member state, Norway from 2019, Iceland from 2018 and the United Kingdom until 2020.
• Models are grouped the way we list them, so trim names such as C 220 d or 320d count towards the C-Class or the 3 Series. Names we cannot match to a model are shown as the EEA records them.
• Fuel shares are of the cars whose fuel was recorded. "Hybrid" covers petrol or diesel cars with an electric motor, plug-in or not, because the registration data does not separate them.
• Average CO₂ is the type-approval value. Until 2017 it is the NEDC figure and from 2018 the WLTP figure; the two tests give different numbers, so they are marked and not compared.

What we never publish

• Owner names, addresses or other personal details; inspector numbers; suspension and unpaid-ticket flags.
• Ratings, reviews or rankings of individual businesses, and no paid placement.