Data methodology and sources

For each make, model and year we compute an odometer distribution and report three points: the 10th percentile (low), the 50th percentile (median, what half of the cars read below) and the 90th percentile (high). The band is the range where most cars of that age fall.

The primary distribution is computed from the UK DVSA anonymised MOT record, which carries an odometer reading at every test but no VIN and no personal data. It is a fleet reference, not a per-car history. The UK record is a large, well-documented reference fleet and is clearly labelled as UK data. Alongside it, where the sample allows, we show the Lithuanian home-market band computed from TRANSEKSTA technical inspections, using each vehicle's latest reading only.

We publish a band only when it rests on at least 1,000 readings. Groups below that threshold are withheld rather than shown thin. The current release covers 471 make, model and year pages, drawn against a Lithuanian fleet set of 2,345 model groups.

Method details for the UK bands. Observation year: 2023, from the DVSA anonymised MOT test results for that year. We keep class 4 tests (cars and light vehicles) with an odometer reading between 1 and 500,000 miles and a usable first-use year from 1995 to 2024; the year shown on a page is that first-use year, not necessarily the manufacturer's model-year designation. Readings are grouped by make, model and first-use year, and a group can include repeat tests of the same car. For groups with more than 12,000 readings, percentiles are estimated from a uniform reservoir sample of up to 12,000 readings drawn with a fixed random seed (20260623), so they are approximate; the published sample size is the number of accepted readings, not the reservoir size. Source readings are in miles; kilometre values are converted at ingestion with a factor of 1.60934 and rounded to whole numbers.

Method details for the Finnish bands. Source: Traficom open vehicle data, extract of 30 Jun 2026 (CC BY 4.0). A band is the 10th, 50th and 90th percentile of the last odometer reading recorded in the register for passenger cars (classes M1 and M1G) in traffic use in Finland on that date, grouped by year of first entry into service, which can have been abroad, so used imports are included. The file carries no reading date; the value is normally the reading from the most recent periodic inspection. Scrapped and exported cars are absent, so older groups describe survivors. Readings are discarded when blank, when they are placeholder values such as 999999, when they are below 1,000 km or above 1,000,000 km, or when they exceed 150,000 km per year of age. First-use years 2022 to 2026 are excluded because fewer than 95% of those cars have a recorded reading yet (they have not been inspected). Groups are stored from 100 vehicles and shown on pages from 1,000. This is a fleet reference, not an estimate for any individual car, and it is not comparable one to one with the UK or Lithuanian bands. The dates differ: the Finnish value is the last reading in a register extract of 2026, while the UK figures are readings from 2023 tests, so cars of the same first-use year are several years older in the Finnish figure, which by itself adds tens of thousands of kilometres. A higher or lower Finnish figure does not show that Finnish cars are driven more or less.

Fuel split. The petrol and diesel rows use the same definition, outlier rules, first-use years and minimum group size as the all-fuels band, but count only cars whose register fuel is petrol or diesel and which are not recorded as hybrids (the fuel and hybrid fields of the register, mapped with the same vocabulary as the VIN-pattern decoder). Hybrid, plug-in hybrid, electric and gas or bi-fuel cars appear only in the all-fuels band. The register leaves the hybrid flag blank for most cars first used up to 2013, so a few unflagged older hybrids may be counted as petrol. In the pre-check, a fuel-specific band is used when the car's fuel is recorded by a registry or, failing that, is the typical fuel of its VIN pattern in the Finnish register (this is stated whenever it applies); otherwise the all-fuels band is used and labelled as mixed. On pages the split appears only where both fuels have at least 1,000 vehicles. Fuel-specific groups stored: 3,214.

Fail rates are initial-test (first attempt, non-retest) failure shares. For the UK proxy we use class 4 MOT tests: a pass is P or PRS, a fail is F, and aborted tests are excluded. Each rate is compared against the median for cars of the same age in the same fleet, so the figure reads as "higher than", "in line with" or "lower than" typical, not as a bare percentage. The Lithuanian first-inspection fail rate is shown from TRANSEKSTA where the sample supports it.

For a single VIN, the free mileage check lines up valid official readings on distinct dates and flags a possible rollback only when a later reading is strictly lower than an earlier one. We do not infer, estimate or guess a reading between records, and we mark gaps between comparable readings. Conflicting values on the same date are shown but excluded from comparisons. Czech inspection exports omit units. Within our imported date range, we treat those values as kilometres for rollback and gap comparisons and label them ‘unit unknown — likely km.’ Affected verdicts disclose this assumption; these readings do not feed the model mileage advisory. Other readings need known units before they enter comparisons. Optional UK plate results are separate: they do not establish that the plate belongs to the submitted VIN. A flag is a prompt to investigate, never a proof of fraud; a clean result is not a guarantee.

Separately, at the model level, we report the fleet rollback prevalence from the Lithuanian inspection record: the share of cars with two or more readings where a later reading falls below an earlier one at some point. Because inspection dates are week-precise and data entry adds noise, we show the larger-than-5,000-km drop share alongside the raw share so the figure is not overstated. This is fleet context, never a verdict on an individual car.

New York DMV inspections: New York only: weekly data covering the last five years. Miles are converted to km. A zero reading may mean a missing or broken odometer and isn’t compared. Entry errors are possible, and no record doesn’t mean a clean history. No title, accident or service history.

Vehicle checks use separate official sources: DVSA MOT searches UK MOT history by registration, and its plate results are not matched to the VIN. ISTP publishes Czech daily inspection exports; our local import covers only the dates shown in the report, so an absent match is not a complete national history. eSyn provides Danish inspection reports for an optional, bounded, user-triggered VIN pilot when connected. Confirm every result against the original report.

Recall signals come from three independent EU-market registers: the EU Safety Gate rapid alert system, the German KBA recall database, and the Dutch RDW open-recall prevalence. We report each source on its own and never add them together: they use different registers and different scopes, so a Safety Gate alert and a KBA campaign may describe the same defect or different ones. The feeds carry no VIN range, so a model-level match is advisory, not a per-VIN applicability claim. Verify any recall with the manufacturer against the specific car.

CO2 and fuel or energy consumption are EU type-approval figures from the European Environment Agency's monitoring of new car registrations under Regulation (EU) 2019/631, reported as a registration-weighted mean per make, model and fuel. Type-approval consumption understates real-world use, so any running-cost estimate built on it should be read as a floor, not a promise.

The free VIN decoder resolves specification from NHTSA vPIC plus an in-house pattern map. The ISO 3780 world manufacturer identifier points to a manufacturer and its assigned country; this is not necessarily the assembly location. We show a factory location only when a source supplies it. We show vehicle facts only, never owner or personal data, and we never reject a European VIN on a US check digit. A model name is emitted only on an exact pattern match; otherwise we show a model family or make and plant and offer an outbound link to a dedicated decoder, rather than assert false precision.

Where a VIN pattern also appears in the Finnish open vehicle register (Traficom, CC BY 4.0), we aggregate the registered vehicles that share it into a table. It can supply a model, plus typical fuel, engine size, power, body and electrification, and the range of years those vehicles first entered service. These describe the pattern, not this car: engine size and power differ for a few percent of vehicles on some makes. When the register disagrees with another decoder’s model, the report says so and names both. Register data is also used to withhold a model year that cannot fit the pattern.

The ninth character is a check digit in North American VINs. European VINs may use that position differently, so a mismatch alone does not invalidate a European VIN. This technical check is not a vehicle-history finding or proof of authenticity.

The report displays available specifications and omits empty fields. Identification, Powertrain and Origin counts show how many fields were returned; they are not confidence scores. A pattern match does not verify this car’s factory options, colour or individual build record.

Fleet distributions are aggregates across many cars, not the history of one car. Model-level recall and CO2 figures describe a model or a production window, not a specific vehicle. A pre-check is a screening tool: it reads what the official record holds and admits what it does not. Inspect the vehicle and its paperwork before buying.

To put this to work, browse typical mileage by model or check one car by VIN for its decoded details and any available vehicle records.