Four in five EU tenders don’t tell you how you’ll be scored

Data note · July 2026

The European Union runs a single, mandatory publication system for public tenders above threshold. Tenders Electronic Daily — TED — is where every above-threshold contract notice in the Union must be published, in a structured, machine-readable form, using a common set of eForms.

Among the fields those forms carry are the award criteria: what the buyer will score you on, and how much each criterion is worth. This is arguably the single most decision-relevant piece of information in the entire notice. It tells a bidder where the points are — and therefore whether the contract is worth a week of their time.

We looked at how often it is actually there.

Across 10,252 open tenders in our corpus, 39% have the award-criteria field populated at all. After removing duplicates, placeholders and non-criteria — described in full below — 19% carry usable, weighted award criteria.

Four in five tenders on the EU's flagship transparency platform do not machine-readably tell you how you will be scored.

The spread is not a rounding error

The headline number conceals something far more interesting. Completion is not distributed. It is national, and the range is close to total.

Country	N	Populated	Usable
Romania	471	100%	80%
Bulgaria	356	100%	16%
Portugal	389	99%	37%
Croatia	249	98%	89%
Latvia	245	98%	48%
Slovakia	199	96%	16%
Czechia	499	94%	29%
Netherlands	679	68%	49%
Denmark	214	59%	41%
Finland	490	47%	15%
Poland	541	46%	26%
Slovenia	217	45%	16%
Germany	777	42%	16%
Austria	218	13%	9%
Luxembourg	112	10%	4%
France	705	7%	4%
Spain	594	3%	3%
Belgium	510	3%	1%
Greece	351	2%	1%
Ireland	317	2%	1%
Italy	608	1%	0%
Sweden	589	1%	1%
Lithuania	334	1%	0%
Hungary	194	0%	0%
Estonia	192	0%	0%
Malta	124	0%	0%
Cyprus	78	0%	0%
All EU	10,252	39%	19%

Romania, Bulgaria, Portugal and Croatia populate the field on essentially every tender.

Hungary, Estonia, Malta and Cyprus populate it on none.

Italy, Sweden, Lithuania, Ireland and Greece are at one or two percent — which, given sample sizes in the hundreds, is not noise. It is a practice.

Same directive. Same forms. Same legal obligation to publish award criteria. And a hundred-fold difference in whether the structured field gets filled.

Populated is not the same as usable

The second finding is subtler, and we think more important.

Filling in a form is not the same as filling it in meaningfully. Look at the two columns together.

Bulgaria populates 100% of its notices. Only 16% survive a basic quality filter.

Croatia populates 98%. Fully 89% survive.

The two countries are indistinguishable on the headline metric and worlds apart in reality. One is filling the field with something a bidder can act on. The other is filling it with something that passes validation.

What ends up in the unusable 84%? Real examples from our corpus:

• Duplicated criteria. The same criterion name repeated once per lot — "Price," "Price," "Price," "Price," "Price," "Price," "Price" — each with a weight of zero.
• Placeholder criteria. One Dutch tender listed a criterion named "Fictief gunningscriterium" — fictional award criterion — with a weight of 1.
• Non-criteria. A German notice used the award-criterion field to state, in prose, that no selection decision would be made.
• Form names instead of criteria. "Prijzenblad" — the name of the price schedule attachment — entered as though it were something you'd be scored on.
• Flat weights. Every criterion assigned a weight of 1, which conveys precisely as much information as leaving them all blank.

Our filter removes these: it deduplicates by criterion name, drops known placeholder and non-criterion strings, drops zero and absent weights, and requires that at least two distinct weighted criteria survive. The full method is below and the filter is a heuristic — a different filter would produce different numbers.

But the direction is not in doubt. The gap between "the field is populated" and "a bidder could use it" is large, and it varies by country in ways the headline number completely hides.

What this means if you're bidding

The practical consequence is simple, and it is not good.

If you are bidding in Sweden, Italy, Ireland, Greece, Belgium, Spain or France, the TED notice will, in the overwhelming majority of cases, tell you nothing about how you will be scored. Not because the buyer is hiding it — the criteria are published, and must be. But they are published in the tender documents, not in the structured notice.

Which means any workflow that begins and ends with the notice — any alert, any feed, any monitoring tool that reads TED and stops there — is showing you a tender with the most decision-relevant information stripped out.

You have to open the documents. There is no way around it. In most of Europe, that is where the award criteria live.

The Netherlands, at 68% populated and 49% usable, is unusually good by EU standards — and it still means half of Dutch tenders don't machine-readably publish their criteria. And our reading of Dutch tender documents suggests that even where TED is populated, the Aanbestedingsleidraad frequently carries a fuller and more precise breakdown: sub-criteria, point allocations, and the scoring formula itself, none of which fit in the structured field.

Method

Corpus. 10,252 open tenders (submission deadline on or after 14 July 2026) from our ingested pool of TED notices.

Populated. The eForms award-criterion fields (award-criterion-name, award-criterion-number) contain a non-empty array.

Usable. After filtering, at least two distinct weighted criteria remain. The filter:

1. Deduplicates criteria by name (removing per-lot repetition, which the eForms structure encourages);
2. Drops known placeholder and non-criterion strings (Fictief gunningscriterium, the German no-selection-decision text, Prijzenblad and similar form names, geen gunning);
3. Drops criteria with zero or absent weights;
4. Requires ≥2 distinct weighted criteria to survive.

Snapshot. A single point in time, open tenders only. Not longitudinal.

Caveats — and these are not decorative

1. This measures the completeness of a form, not the conduct of procurement. A country at 0% has not stopped publishing award criteria. Its buyers are not populating the structured eForm fields. The criteria are in the tender documents, where they are legally required to be. This is a finding about the data layer, not about transparency in the substantive sense. Do not read it as "Hungary hides its award criteria." Read it as "Hungary's buyers don't fill in that box."
2. "Usable" is our operationalisation. The junk filter is a heuristic. It is published above and it is reproducible, but a different filter yields different numbers. The populated column is objective; the usable column is a judgment we have tried to make transparent rather than authoritative.
3. The corpus is our ingested pool, not a census of TED. Our ingestion is CPV- and country-driven, so the per-country N values are not national tender counts and should not be read as such. The rate within each country's sample is defensible — there is no plausible reason why our ingestion criteria would correlate with whether a buyer fills in the award-criteria field — but the absolute numbers are a sample, not a population.
4. Small samples at the bottom. Cyprus (N=78), Luxembourg (N=112) and Malta (N=124) are the smallest. Their rates are directionally clear (0-10%) but should be read with the sample size in mind.
5. eForms adoption is moving. The eForms regime is relatively new and national practice is still settling. These numbers could look quite different in two years — which is itself an argument for measuring them again.

Why we ran this

We build software that reads tender documents. So we have an obvious interest in the finding that the notices are incomplete, and you should weigh that accordingly.

We are publishing the method, the filter, and the caveats precisely so you don't have to take our word for it. And we would be glad to be corrected: if someone has a better filter, or a fuller corpus, or a reason to think these numbers are wrong, we would like to know.

What we can say is that the question — how often does the EU's mandatory publication system actually tell a bidder how they'll be scored? — appears not to have been asked. It is answerable, it took an afternoon, and the answer is: less often than you'd think, and it depends enormously on where you are.

---

This note is part of a series examining what is and isn't measured in European public procurement. A companion study, currently in progress, examines the price-scoring formulas used in Dutch tenders — and how many of them expose bidders to rank reversal.

Corrections, questions and better data are welcome.