How to automate EU tender search — and what it actually gets you

Why searching by hand systematically loses you contracts

Procurement notices are not published in one place. They are spread across the EU journal, national portals, regional portals and commercial e-procurement platforms — in 24 languages. Checking them by hand costs you three times over.

• Time: checking each source individually consumes working hours that are missing later, when the bid has to be written.
• Noise: broad keyword alerts return the wrong trade, the wrong country and contract sizes that do not fit your business.
• Deadlines: the later a suitable procedure surfaces, the less time is left for documents, evidence, partners and pricing. Open procedures above the EU thresholds have a minimum time limit of 35 days — losing two weeks of that means a weak bid or none at all.

What an automated search technically does

The first step is scheduled polling: portals are queried through their APIs and data services, not opened by hand. The second step is classification. The EU maintains a common vocabulary for exactly this purpose — the Common Procurement Vocabulary (CPV). Every notice carries at least one CPV code that classifies the service or product, independent of the language it was written in.

The third step is matching against the company profile. What counts is not just whether a keyword appears, but whether place of performance, contract value and deadline fit your business and whether any exclusion criterion applies. Each of those checks contributes to a score that can be broken down — you see which criterion contributed what.

What to define when you set up a profile

The quality of the matches depends almost entirely on the care put into the profile. Four inputs matter most:

• Trades and services, ideally phrased the way contracting authorities phrase them — and backed by CPV codes, because codes work across all 24 EU languages while keywords do not.
• Exclusion terms for services you explicitly do not offer. They save more time than any additional keyword.
• Place of performance and radius. A contract 400 kilometres away is not a match, even if the trade fits.
• Contract value from/to. Lots that are too large burn estimating time with no prospect; lots that are too small are not worth the overhead.

How to recognise a solution you can trust

The decisive test is the justification. A service that reports a match without saying why cannot be corrected: you can neither sharpen the profile nor judge whether a gap is the tool’s fault or your input’s.

The second test is source transparency. Every match should link to the original notice, and the service should publish which sources it actually polls and when each last succeeded. Without that status you cannot know whether a source has been silently failing for days — our own source status page and /api/source-status endpoint exist for exactly that reason.