Methodology and data source

What data this uses, how it is processed, and what to be careful about.

Source: TED, the EU's official procurement journal

All notice data comes from Tenders Electronic Daily (TED), published by the Publications Office of the European Union, through its public, anonymous Search API (POST https://api.ted.europa.eu/v3/notices/search). We do not scrape the website. The pipeline is designed to stay well inside TED's published fair-use limits (700 requests per minute): it needs only a handful of requests, once a day.

According to TED's legal notice, unless otherwise noted the procurement notices published in the Supplement to the Official Journal can be freely reused for commercial or non-commercial purposes. We link to the official notice from every brief and do not use EU or TED logos or emblems.

No affiliation

This project is independent. It is not affiliated with, endorsed by or connected to the European Union, the European Commission, the Publications Office of the European Union or TED.

What we select

• Notice type: contract notices (calls for bids). Award notices and others are excluded.
• Still open: at least one lot with a deadline today or later. For two-stage procedures (restricted, negotiated, competitive dialogue) TED usually publishes only the deadline to apply; the brief labels it as such.
• Sector: IT & digital services: notices whose primary CPV classification starts with 72 (IT services) or 48 (software packages and information systems). Matching on any CPV code in any lot would include notices where software is only a small part (about 21% of IT-matched notices in our sample), so we use the primary classification.

How a brief is made

• Title, buyer, country, deadline, CPV code, procedure type and accepted bid languages are copied from TED fields. The title and a short description are shown in English. TED notices are written in the buyers' own languages, so: if the notice is in English, or the buyer supplied an English title, we use that text as written; otherwise the title and the first sentences of the description are machine-translated into English by an AI language model and labelled “Machine-translated from [language] — check the original”. Next to every translation we show the original title and a link to the original TED notice, which is the authoritative text. Machine translations can contain mistakes; names and numbers should always be checked against the original. If no translation is available the brief shows the original text and says so. The sample briefs on this site were translated once by an AI model and are stored in a data file marked as machine-translated; translating live briefs needs a configured LLM provider and costs money per notice, so it is off until one is set up. Translating briefs into Swedish, German or French is a possible later feature.
• Value is shown only if the notice states one (procedure-level estimate, or the lot estimates when there is no procedure-level one, labelled as such). All values are TED's "estimated value excluding VAT"; we do not convert currencies. If there is none, the brief says "Not stated in the notice".
• Fit hints are simple rules (days left, contract size bands, procedure type, accepted languages). They are heuristics, not eligibility checks. A small firm may still be excluded by turnover, reference or certification requirements in the tender documents; the “Can a small IT firm bid?” flags below show only what the notice itself states about them.
• Origin label: under every brief a one-line label says whether the text shown is the “Original text in English” or “Translated from [language]”. It comes from the same data as the translation note above (the notice language and how the English text was obtained). Where the buyer supplied the English title in a notice written in another language the label adds that, and where no English text exists it says “Not translated”.
• Summary: either an automatic template that restates TED fields (and the English title/description above), or, when enabled, a short AI-written summary instructed to use only the notice text. The brief states which one was used. AI summaries can contain mistakes, which is why the official link is always included.

“Can a small IT firm bid?” flags

Each brief has a short list of plain facts about things that often decide whether a small firm can take part. They are read from structured fields of the TED notice (the eForms data), one line each. They are not a score, advice or an eligibility check, and nothing is guessed from free text. When the notice has no data for a line it says not stated in notice; when the fields were not fetched for an old stored brief, that brief shows no flags at all rather than a misleading “not stated”.

• Contract form: framework agreement (with or without reopening of competition) or dynamic purchasing system, from the notice's own fields. In our sample of 240 IT notices, 19% state a framework agreement and 0% a dynamic purchasing system.
• Lots: number of lots and any limit on how many lots one bidder may bid for or win (13% of the sample has more than one lot).
• Turnover, references, certifications: shown only if the notice itself lists a selection criterion of that type (eForms criterion codes, for example “average yearly turnover”, “references on specified services”, “independent quality assurance certificates”). We do not extract amounts, years or numbers of references; where the notice has criterion text we say so and point to it. Many notices do not list criteria at all: in the sample, 37% list at least one selection criterion in the notice and 63% say their criteria are only in the procurement documents (or the ESPD form). For those the flag says “not stated in notice”, which means read the tender documents, not “no requirement”. Some notices use an older, coarser list (for example “economic and financial standing”) that does not say whether it means turnover; we show it as exactly that.
• Security clearance: as stated in the notice's yes/no field.
• Language of bids, deadline length, estimated value: the languages the notice lists, the number of days from publication to the deadline (to apply, for two-stage procedures), and the stated value (EUR or other currency, never converted; a framework's maximum value is added when it differs).
• Open purchasing systems: when the notice's buyer is Region Stockholm or Adda Inköpscentral and the notice says it uses a dynamic purchasing system, the brief adds a pointer to that buyer's public page about its open systems. Only these two buyers are covered, because they are the only ones we have checked and linked in our research (desk research of 17 October 2026; not legal or procurement advice); the pointer says it is not known which system the notice belongs to. The Adda page was not checked for IT consultancy coverage. Other open systems (and frameworks, which are closed to new suppliers) get no pointer.

Limits: the flags only see what the notice's data fields contain. They cannot see the tender documents, where most Swedish and many other requirements are written, nor direct awards or framework call-offs, which are not advertised as notices. Fields can be wrong or inconsistent (for example a placeholder value of 1 in some dynamic-purchasing-system notices), and we show them as stated. The original notice on TED always prevails.

Fit score and short list

Each notice also gets a fit score from 0 to 100 for a firm profile. The score is calculated by fixed rules in plain code (src/fit.js). It does not use AI, learns nothing from data, and gives the same result for the same notice, profile and date. It is a sorting aid for building a short list, not a prediction that you will win, not an eligibility check, and not a recommendation to bid. The weights below are our own assumptions, not measured facts.

The default profile is a small IT/software consultancy, EU: main CPV codes 7220, 7221, 7222, 7223 and 7226 (software programming, consultancy, development and support), 48 (software packages) and other 72 (IT services); buyers in any of the 27 EU member states; bids accepted in English; at least 7 days left, ideally 14 to 60; estimated value EUR 20,000 to 1,000,000. A different profile changes the score and the order.

Notices are ranked by score; ties go to the earlier deadline, then the lower publication number. The short list is the top 10 notices of a run with a score of at least 40. Every entry shows the points and the reason per factor, so you can see why it is where it is.

Limits: the score cannot see whether the contract is mainly goods or licences rather than services, whether turnover, reference or certification requirements exclude a small firm, or whether the text is a good match beyond the keyword list; notices that state no value or no language get partial points rather than being ranked on facts we do not have. English keywords only match English text, so a notice written in another language and not yet translated may score low on keywords. At the moment the score and short list exist in the code and in the text rendering of a brief; the public sample page shows no scores.

Known limits

• Roughly half of the IT notices we sampled (48%) state no value.
• TED carries notices above the EU procurement thresholds. Smaller national tenders may only be on national portals and are not covered.
• Notices can be corrected or cancelled after publication. A brief reflects the data when it was generated.
• TED's own disclaimer applies: the Publications Office accepts no liability for the information, and only notices published in the Official Journal Supplement are authentic.

Reporting an error

Operated by monocube AB, Elbevägen 48, 141 32 Huddinge, Sweden. Corrections and error reports: contact@tenderbrief.eu (this is a prototype and replies are not guaranteed).