Why TED alone will mislead you: the hidden problems in EU procurement data

We've spent the past several months building TedScout — connecting to 13 national procurement portals, ingesting 8 million award notices, and writing the code that normalises, deduplicates, and queries this data. In the process we found nine specific ways that TED data misleads the people who use it. Some are TED's own limitations. Some are inherent to how procurement data gets published across 27 different legal systems. All of them have real consequences for suppliers trying to price bids, target markets, or understand who's winning contracts.

---

1. Framework agreements inflate every contract value benchmark

When a government department sets up a 4-year IT framework agreement with a ceiling value of €40 million, TED publishes that as a single contract award notice worth €40 million. The individual call-off contracts — the actual work orders placed against that framework, typically €200k–€500k each — are usually not published individually.

The result: any analysis of TED award values is systematically skewed upward. A supplier trying to benchmark IT contract values in Poland from TED alone would find a median of €1.83 million. The same analysis using BZP — Poland's national portal, which publishes individual award notices for actual work — gives a median of €184k. The same market, the same contracts, a 10× difference in apparent value.

This isn't a TED bug. It's how framework procurement is supposed to work — you publish the framework, not every call-off. But it makes TED award data unreliable for pricing benchmarks without significant adjustment.

The EU's own Publications Office identified this as a documented data quality issue. Their TEDDQAIS (TED Data Quality Auditing and Improvement Solution) programme specifically flags incorrect contract values — including one notice with a dummy value exceeding €9 trillion — as requiring systematic cleaning before the data can be used analytically.

2. TED captures only a fraction of actual procurement activity

TED publishes contracts above the EU thresholds: €143,000 for central government services, €5.5 million for works contracts. Everything below those values is published on national portals only, following national rules that vary significantly between member states.

The practical effect is dramatic. Italy's ANAC database — which captures the full Italian market including sub-threshold contracts — contains 237,000 construction contracts since 2023. TED has 1,369 Italian construction awards for the same period. That's less than 1% of the actual market.

External estimates suggest sub-threshold contracts represent 60–70% of all public procurement by number of contracts. For suppliers targeting smaller, more accessible opportunities — maintenance work, IT support, consultancy engagements — TED is showing them less than a third of what's actually available.

3. Award notices are legally required but inconsistently published

EU directives require a contract award notice to be published within 30 days of contract signature. In practice, compliance is patchy.

Finland's State Treasury conducted the most rigorous analysis we've seen: in 2025, award notices were left unpublished for approximately 40% of procurement procedures above the EU threshold. Their conclusion was direct — the information base for understanding public procurement is fundamentally incomplete, even for contracts that are legally required to be published.

"A procurement data repository does not emerge by chance."
— Juho Savonen, Programme Manager, Valtiokonttori (Finnish State Treasury), 2025

Finland is not an outlier. The same structural problem — buyers who run the procurement but have no strong incentive to complete the award notice paperwork — exists across the EU. The data gap is not random either. Direct awards and negotiated procedures are particularly likely to go unreported, which skews any competition analysis drawn from TED data.

4. CPV codes are self-declared and unvalidated

CPV (Common Procurement Vocabulary) codes are the classification system that lets suppliers filter procurement by sector. An IT services contract should be classified under CPV 72; a construction contract under CPV 45.

In practice, CPV classification is done by the buyer and is never validated. A software development contract might appear under CPV 72 (IT Services), CPV 48 (Software packages), CPV 79 (Business services), or even CPV 73 (Research and development) depending on which procurement officer filled in the form and how they interpreted their own project.

The EU's own data quality programme lists CPV accuracy — "who bought what" — as one of its four primary data quality challenges. Any supplier relying on CPV filters to find relevant opportunities will miss a significant fraction of contracts that were miscoded at source.

5. National portal data is not translated

TED provides machine-translated PDFs for every notice in all 24 EU official languages — this is genuinely useful for above-threshold contracts. However, the national portals where the majority of procurement actually happens are a different story.

ANAC in Italy publishes in Italian. SEAP in Romania publishes in Romanian. BZP in Poland publishes in Polish. BASE in Portugal publishes in Portuguese. None of these portals offer machine translation. A German IT company monitoring Italian procurement for opportunities will find the contract notices, but reading and interpreting the specifications, award criteria, and technical requirements requires either fluency or a translation process that doesn't exist in most procurement tools.

This is less a TED limitation than a structural feature of how national procurement works — each portal serves its domestic market and has no mandate to publish in other languages. But any tool that claims to cover "EU procurement" without addressing the language barrier is giving suppliers a false sense of coverage.

6. France has over 300 procurement platforms

Under the French Public Procurement Code, buyers are required to publish above-threshold contracts to BOAMP or TED. Below the €90,000 threshold, they can publish exclusively on their chosen buyer profile platform — a commercial or regional procurement portal of their selection.

The result is a fragmented ecosystem of over 300 platforms hosting French below-threshold tenders: AWS-Achat, Marchés Sécurisés, e-marchespublics, Marchés Online, Maximilien (the Île-de-France portal), Megalis Bretagne, and dozens of regional equivalents. A tender published on Maximilien for a Paris municipality does not appear on BOAMP, does not appear on TED, and is invisible to any tool that monitors only the official channels.

France has partially addressed this through DECP (Données Essentielles de la Commande Publique) — a mandatory open data publication requirement that aggregates award notices from all public buyers to data.gouv.fr. Since January 2024 this is legally required. The DECP dataset contains 3 million French contracts. BOAMP, the source most procurement tools use for France, contains around 41,000.

7. Winner names are free text with no entity resolution

TED records the winner of each contract as a free-text field. The same company appears in the data as "Accenture", "Accenture SL", "Accenture Ireland Limited", "ACCENTURE", "Accenture Solutions S.L.U.", and variants depending on which legal entity submitted the tender and how the contracting authority wrote it up.

The EU's own TEDDQAIS programme lists "name harmonization" as one of its core data improvement tasks — the Publications Office's solution requires building self-evolving master data that links variant names to canonical entities. Without this, any analysis of market share, incumbent identification, or competitive landscape drawn from raw TED data is systematically incomplete.

A supplier asking "how many contracts has Capgemini won in Germany in the last three years" cannot get an accurate answer from raw TED data. The answer requires entity resolution that TED doesn't provide and most procurement tools don't attempt.

8. TED is mid-transition between two incompatible data schemas

In October 2023, TED completed its migration from legacy notice forms to eForms — a new standardised XML schema defined by Implementing Regulation (EU) 2019/1780. This was the largest structural change to EU procurement data in over a decade.

For anyone building on TED data, this means the historical archive and recent notices use fundamentally different data structures. Fields that existed in the legacy schema don't always have direct equivalents in eForms. New eForms fields — structured lot information, detailed award criteria weighting, machine-readable sustainability fields — have no equivalent in legacy notices.

The EU continues to iterate on the eForms SDK, with new versions released roughly every six months. Any tool built on TED data is either maintaining two parsers in parallel, or has a gap in historical coverage. The transition is necessary and ultimately beneficial — eForms data is richer and better structured — but the migration period creates real complexity for anyone who needs consistent data across time periods.

9. Duplicate notices distort the picture

TED publishes corrigendum notices — corrections to previously published notices. It also publishes contract modification notices when awarded contracts are amended. And when framework agreements are revised or extended, a new award notice is published reflecting the updated value.

The result is systematic duplication. In our own dataset, we found construction contracts appearing two to six times with identical values but different publication dates — framework agreement revisions being stored as separate award records. A naive count of contracts in affected markets would suggest a larger, higher-value market than actually exists.

The TED dashboard explicitly documents this: it distinguishes between "duplicates" (earlier versions or updates) and "non-duplicates" (most recent version) and notes that substituted values are generated using heuristic rules whose accuracy cannot be guaranteed. De-duplicating TED data correctly requires understanding the notice lifecycle, not just querying the database.

---

What this means in practice

None of this means TED is useless. For monitoring above-threshold open tenders across the EU, it remains the authoritative source. The eForms migration is making the data progressively better structured. The machine translation capability is genuinely helpful.

But a supplier trying to answer real business development questions — what do comparable contracts actually cost, how competitive is this market, who are the incumbents, what opportunities exist below the headline threshold — cannot get reliable answers from TED alone.

Across Europe the data they need is distributed across hundreds of national, regional, and local procurement portals — published in 24 official languages, structured in incompatible schemas, partially de-duplicated, and almost entirely untranslated. France alone has over 300 buyer profile platforms. Assembling it into something queryable requires solving the problems listed above, not pretending they don't exist.

That's what we've been building. TedScout currently connects to 13 portals — the highest-volume sources across 9 countries — normalises 8 million award notices into a consistent schema, applies per-source data quality rules, resolves entity variants, and makes the result queryable in plain English — inside Claude, Mistral, Cursor, or directly at tedscout.eu. We're adding portals continuously.

EU procurement data is inherently messy — that's the entire argument of this article. Anyone presenting it as clean and unified is either not looking closely enough or not being straight with you. What we can do is be explicit about what each source actually contains, apply consistent quality rules, and tell you where the numbers are solid and where they need context.

If EU public sector is part of your growth strategy, you should know what the data you're using actually contains — and what it's missing.