The register exists, and it can only answer questions you already know

A company that has just started importing needs a forwarder and a customs broker before its first container moves. The obvious candidate signal is the EORI number, the customs registration every EU importer and exporter must hold. New EORI, new importer.

It does not work, and the reason is worth writing down because no amount of persistence would have fixed it.

We pulled the service definition and made live calls against it. The service exposes exactly one operation: validate a number you already have. It takes EORI numbers as input and returns a status.

The response has no registration date field. Not withheld, not gated: absent from the schema. So even with the complete register in front of you, there is no way to ask who registered recently. The data required for the signal is not modelled.

Name and address are consent-gated on top of that. Of five valid numbers we probed, one returned a name.

This is the useful shape of a negative result. A source can be free, public, official and completely unable to answer your question, and you find that out from the schema rather than from the documentation.

Planning applications for warehousing. When a company applies to build or occupy storage and distribution space, that is public, and it happens well before the logistics contracts are signed.

The UK has a free planning API with daily updates that filters cleanly on the storage and distribution use class. The Netherlands publishes municipal notices. Both are real, both are queryable, and this is where the vertical earns its “workable, narrowly”.

Every free national planning source we tested withholds the applicant's name.

The UK API returns the literal string "See source" in the applicant field on every record. Dutch municipal notices refer to “de aanvrager”, the applicant, and give only an address. The Zurich open dataset states plainly that names have been deliberately removed.

So you can find the building for nothing and cannot find the company without work. That inverts the usual assumption that finding candidates is the hard part. Here the candidate list is free and the identification is the entire cost, which changes what the pipeline should look like: a cheap wide sweep, then paid or manual resolution on a small shortlist.

Keyword-matching planning applications for warehousing surfaces a lot of entries reading “change of use from storage and distribution to general industrial”. Every warehouse keyword matches. It is an anti-lead: a building being converted out of storage, by an owner with less use for a logistics provider than before.

Direction matters more than presence. A filter that matches the words will rank your worst prospects at the top.

Two countries. We found nothing comparable for Germany, Austria or Switzerland on the warehousing route, and the customs route is closed everywhere.

Several sources refused inspection, so their claims are unchecked. A US import-data service returned 403 to every request. The German federal employment agency's job API rejected both key and OAuth attempts.

That second one matters. Job advertisements name the employer every planning source withholds, and it is worth retrying from a German address.

We also could not establish how many UK authorities actually contribute, so coverage is unknown.