Methodology

P10 P25 Median P75 P90

• Median half of the comparable cars are listed below it, half above.
• Typical range the middle half of asking prices, from the 25th to the 75th percentile. Not a confidence interval.
• P10–P90 the middle 80%, shown from 30 cars.

What the figures are

• Asking prices advertised by professional sellers (dealers), not transaction prices. Private sellers are not included.
• Gross prices: what a private buyer pays the dealer, VAT included where it applies.
• Used cars only. New cars, pre-registered and demonstration cars (km 0), commercial vehicles, campers, damaged cars and classics are excluded.
• Current figures use the cars that were on sale at some point in the last 10 days.

Sources

Dealer stock published on online marketplaces and on dealers’ own websites in Germany, Austria, Belgium, Spain, France, Italy, the Netherlands and Portugal. We name sources by type, not by site. The share of each type per country is in the coverage table below. Coverage varies by country; see the dated coverage table below.

Cleaning

• Make and model. Free-text make and model names are mapped to a reviewed list of makes and model families (for example, “C 220 d” and “C-Klasse” both become the Mercedes-Benz C-Class).
• Fuel, gearbox and body. Labels in five languages are mapped to one fixed vocabulary. Unknown values stay unknown: such a car still counts in figures that do not use that attribute.
• Price basis. Italian marketplace listings whose price equals their stated net price (about 6% of Italian cars) are left out: we cannot tell whether VAT is included.
• Validity rules. Prices below €500 or above €5 million, monthly lease rates shown as prices, placeholder mileages and more than 100,000 km a year are removed.
• Outliers. Within a group of at least 20 comparable cars, prices far outside the group’s range (more than 4 standard deviations on a log scale) are removed.

Duplicates

The same car is often advertised more than once: on several sites, or twice by the same dealer. It is counted once. Matching uses identifiers where a listing carries them, and otherwise a fingerprint of make, model, year, mileage and price. Identifiers are used only during processing and are never published.

Publishing rules (disclosure control)

A figure is published only if all of these hold:

Rule	Threshold
Cars in the group	at least 5
Different dealers	at least 3
Share of the largest dealer	at most 60%

Within a published group, the median is shown from 5 cars, the middle half (P25–P75) from 10 cars and P10–P90 from 30 cars. Groups that fail are left out entirely, never shown as zero. Asking-price distributions use medians and percentiles. Counts and average asking-price changes are also published where the disclosure rules permit. Prices are rounded to the nearest €100 (€50 below €10,000).

VAT

The headline figure is the gross asking price. For business buyers we also show the share of cars sold VAT-deductible (among cars whose VAT status is known) and their median net price. Portugal: the VAT split is not published yet, because one source labels it unreliably; a fix is in progress.

Gross prices are not directly comparable across countries. They include VAT (19% in Germany; 20% in Austria and France; 21% in Belgium, Spain and the Netherlands; 22% in Italy; 23% in Portugal) and, in some countries, a registration tax already paid on the car (for example ISV in Portugal, IEDMT in Spain and BPM in the Netherlands). That is why every country comparison sits next to the landed-cost calculator.

Like-for-like country index

A simple comparison of medians mixes up price with the type of cars on sale: one country may simply have newer cars. The index compares like with like:

1. Cars are grouped into strata of model year × mileage band × fuel.
2. A stratum is used when the eight-country pool and at least two countries have a published median.
3. For each country: index = Σ weight × (country median ÷ eight-country median) × 100, where each stratum’s weight is its share of the pooled cars, re-scaled over the strata that country has.

100 is the eight-country level; 91 means the same car is listed about 9% cheaper. The number of strata behind each index is shown with it.

How a comparison is chosen

Every figure for a particular car, on a model page with filters, in the listing check and in the calculator, comes from one lookup. It always says which cars it used: model, country, registration years, fuel, mileage band, the number of cars and dealers, and the data date.

1. First the exact group: the same registration year, fuel and mileage band.
2. If that group is too small to publish, the years widen to ±1 around the car’s year, keeping the fuel and mileage band.
3. Then the mileage band is dropped, then the fuel, and finally all years are used.

Each step that was needed is shown, with how far it moved (for example “Years: 2019 → 2018–2020 (±1)”) and the narrower groups that were too small. The exact number of cars in a group that was too small is never shown: that is the point of the publishing rules.

Year ranges. You can choose registration years from–to. When a range matches a published group exactly (±1 year around one year, with fuel and mileage band), that group is used. Otherwise the figure is pooled from the published single-year groups inside the range, each year weighted by its number of cars, and is marked ≈:

• the number of cars is exact; the median and the middle half are estimates; median mileage, median age and the VAT split are not estimated;
• the number of dealers is a lower bound, because one dealer can sell cars of several years;
• years without a published figure are left out and named, and a range needs a published figure in at least half of its years;
• where most of the cars sit in years with fewer than 10 cars, the estimate is marked as rough.

We tested the pooling on the data of 2 November 2026 against 17,485 published three-year groups whose single years were all published: the pooled median was within 1.2% of the exact median for half of them and within 5.1% for nine in ten (0.9% and 3.3% when at least 80% of the cars sit in years that publish the middle half).

Trends

A market-wide trend adjusted for changes in stock composition is not yet published. We separately publish weekly asking-price changes for cars that remain advertised; see the price-outlook methodology.

Pages for thin markets

A model page for one country is shown to search engines only when it has at least 30 cars and at least 3 published model years. Smaller pages stay available, marked as thin.

Limitations

• Asking prices are not selling prices; discounts are not visible.
• Dealers only: private sales are not covered.
• Coverage differs between countries and changes as sources are added.
• A small share of cars may be assigned to the wrong model family; the quality table below tracks how many names we map.

Versions

The methodology has a version number. A change of the first number means the figures are not comparable with earlier ones.

• 1.0.0: first version.

Current thresholds

Minimum cars per figure	5
Minimum dealers per figure	3
Largest dealer’s share, at most	60%
P25–P75 from	10 cars
P10–P90 from	30 cars
Cars on sale within	10 days

Coverage

Country	Cars	Marketplaces	Dealer sites
Austria	42,020	97%	4%
Belgium	2,902	46%	54%
Germany	105,729	91%	9%
Spain	14,084	48%	52%
France	21,260	–	100%
Italy	103,370	72%	29%
Netherlands	109,915	49%	51%
Portugal	32,335	96%	4%

Data quality

Measure	Value	Target
Canonical make	99.8%	99.5%
Canonical model family	99.1%	97%
Fuel normalised (of non-empty)	100%	99%
Gearbox normalised (of non-empty)	100%	98%
Body type known	98.5%	90%
VAT status known	40.8%	–
Duplicates removed	3%	–
Groups suppressed by the publishing rules	64.3%	–