Traceable to the publication it came from.
Most European registration data is public. The difficult part is being able to defend a figure when someone challenges it in your report, and that is what this page covers.
From publication to delivery
Acquisition
Series are retrieved from the publishing authority itself, not from an intermediary. Retrieval date, file version and publication reference are recorded with the data at the moment of download.
Harmonisation
Manufacturer, brand, model and powertrain taxonomies are mapped to a single internal reference so that a marque means the same thing in Warsaw and in Lisbon. Mappings are versioned and the mapping table travels with the delivery.
Validation
Compiled totals are reconciled against the aggregate the authority itself published. Divergences are not smoothed away. We flag them, quantify them and explain them in the release note.
Delivery
Structured files or feeds with a source manifest. Restatements go out as a new version with a changelog, not as an overwrite of a series you have already used.
Why we publish our sources
Across the major providers of European automotive data, we could not find one that names the registries it compiles from. The largest describes "a wide range of sources". Another refers to "partnerships with ministries and official data sources". The intelligence arm of the manufacturers' own association names none at all.
We do the opposite, and for a practical reason. An institutional buyer sooner or later has to answer a question about provenance, from a peer reviewer, an auditor, a regulator or a journalist. A supplier who cannot answer it has simply pushed the problem onto the client. The full source map is on the data coverage page.
What we do not do
We do not resell, repackage or redistribute third-party commercial databases. We compile from primary public sources published under terms that permit reuse, and we observe the licence conditions of each: CC BY 4.0 for the European Environment Agency, the Italian ACI, Finnish Traficom and the Polish CEPiK API; CC0 for the Dutch RDW; the Open Government Licence v3.0 for United Kingdom vehicle licensing statistics. Where a source restricts redistribution, as ACEA's monthly releases do, we use it only to cross-check and we say so.
We also do not present modelled figures as observed ones. Where a series is estimated, interpolated or nowcast, it is labelled in the data, not only in the documentation.
Access constraints we work around
Public does not mean frictionless, and the constraints move. Since 1 February 2025, the Spanish DGT's registration microdata files no longer carry the full vehicle identification number; obtaining it requires a substantiated legitimate-interest application. Several authorities publish provisional figures that get revised weeks later. Austrian registrations, for instance, are provisional around the tenth of the month and definitive by mid-month. Our release notes state which vintage a delivery reflects.
Regulatory context
Two instruments shape most of the questions our data is used to answer.
Regulation (EU) 2019/631 sets CO2 performance standards for new cars and vans. Under Article 7, each Member State records the prescribed information for every new vehicle registered on its territory and transmits it to the Commission by 28 February for the preceding calendar year. That is why the European Environment Agency dataset is as much a registration dataset as an emissions one. The targets are a 15% reduction against 2021 for 2025–2029 (93.6 g CO2/km for cars, 153.9 g/km for vans), 49.5 g/km and 90.6 g/km for 2030–2034, and zero from 2035.
Regulation (EU) 2025/1214, adopted on 17 June 2025, changed how those targets are assessed for 2025, 2026 and 2027: compliance is now measured on the three-year average of a manufacturer's specific emissions rather than year by year. Any analysis that reads the 2025 or 2026 figures as a standalone compliance test is reading them wrongly, and we flag this in the documentation of the affected series.
Revision policy
National authorities revise. When a source restates a period we have already delivered, we issue a new dataset version, retain the superseded version, and document what changed and by how much. Clients who have published analysis on an earlier vintage can therefore reproduce exactly what they published.