EU public procurement competition: 519,004 lots
One bid is the ordinary outcome in EU public procurement competition. Across 519,004 award lots published in the EU official journal between March and August 2026, 36.1 per cent drew exactly one bid. Inside the euro area the rate is 29.2 per cent, and 49.2 per cent of lots there drew two bids or fewer. Just under a third of all priceable euro-area contract value, EUR 61.2bn, went on a lot where only one firm bid. This article is for bid managers, procurement officers, public bodies, journalists and analysts who need the counts, the method and the limits behind that story.
What we actually measured
The source is TED, the Tenders Electronic Daily supplement of the Official Journal of the European Union. We loaded the monthly eForms bulk packages for March, April, May, June, July and August 2026 into a single SQLite database, the same build that backs the published European public procurement competition study. This article adds 25 cuts to that build. It does not re-scrape the source or repeat the extraction.
The funnel, in the order the filters fire:
| Step | Notices | Lots |
|---|---|---|
| All notices in the March to August 2026 packages | 461,720 | 761,062 |
| Rows that carry a bid count | 149,857 | 584,516 |
| Lot rows recording zero bids (held out) | 0 | 65,339 |
| Lot rows claiming more than 200 bids (held out) | 0 | 73 |
| Lots in the main competition cut (1 to 200 bids) | 149,857 | 519,004 |
| Lots inside the euro area | 0 | 172,978 |
| Euro-area lots with a usable euro value (EUR 1k to EUR 500m) | 0 | 122,032 |
Four rules govern everything below. A lot enters the competition cut only if it reports a bid count between 1 and 200, so single-bid means exactly one bid and not a missing field. Money work is restricted to euro-area countries and to lot values between EUR 1,000 and EUR 500m, with no currency conversion. The 2,949 euro-area lots above EUR 500m and 56,732 below EUR 1,000 are excluded and reported separately, not deleted. Nothing here is a trend, because the packages carry a load stamp in the year field, so a "year" comparison would measure when we downloaded the file and not when the buyer ran the tender. The window is six months of publication dates, and awards published in July 2026 can relate to tenders run in 2024.
The cut covers 149,857 distinct award notices from 26,490 buying organisations across 58 buyer countries. Buyers exist here only as the name string typed into the notice, with no company register behind it, so two spellings of one hospital count as two buyers. That limits the buyer-level cuts, which is why several are reported twice, once grouped by buyer name alone and once by buyer name plus country.
One bid is the ordinary outcome in three euro-area lots in ten
Of the 172,978 euro-area lots in the cut, 50,578 attracted exactly one bid, which is 29.2 per cent. A further 34,463 lots attracted two bids, so 49.2 per cent of euro-area lots drew two firms or fewer. The median euro-area lot drew three bids.
Pull in the countries outside the euro and the rate climbs: 187,430 of 519,004 lots, or 36.1 per cent, on a median of two bids, with 55.7 per cent of lots at two bids or fewer. The gap between 36.1 per cent and 29.2 per cent is mostly about which countries dominate the lot counts. Poland and Romania together supply 294,230 of the 519,004 lots, and both sit above the euro-area average.
| Bids recorded on the lot | All countries, lots | Share | Euro area, lots | Share |
|---|---|---|---|---|
| 1 | 187,430 | 36.1% | 50,578 | 29.2% |
| 2 | 101,831 | 19.6% | 34,463 | 19.9% |
| 3 | 75,374 | 14.5% | 26,144 | 15.1% |
| 4 | 53,300 | 10.3% | 18,055 | 10.4% |
| 5 | 34,459 | 6.6% | 11,814 | 6.8% |
| 6 to 10 | 51,078 | 9.8% | 21,233 | 12.3% |
| 11 to 20 | 11,478 | 2.2% | 7,631 | 4.4% |
| 21 to 50 | 2,862 | 0.6% | 2,007 | 1.2% |
| 51 to 200 | 1,192 | 0.2% | 1,053 | 0.6% |
The shape of the table is the finding. One bid is the largest bucket in both cuts, and the distribution falls away after it. Deep competition is rare: 3,060 euro-area lots, 1.8 per cent of the total, saw more than twenty bids.
One caution about the word "competition". A bid count is whatever number the buyer typed into the award notice. It does not say how many firms downloaded the documents, how many were capable of bidding, or whether the one bid was priced against a credible alternative, and it says nothing about price. The figures below describe the number of firms that reached the point of submitting a tender, the only contest measure the official data carries.
Country by country, with the lot counts behind each rate
Rates without denominators are little use, so the table below gives the lot count, the single-bid rate, the share of lots at two bids or fewer, the mean bid count and the median lot value for the twenty euro-area countries with at least 300 lots in the cut.
| Country | Lots | One bid | One bid % | Two or fewer % | Mean bids | Median lot, EUR |
|---|---|---|---|---|---|---|
| Cyprus | 449 | 249 | 55.5% | 74.6% | 2.07 | 247,014 |
| Greece | 5,280 | 2,835 | 53.7% | 72.5% | 3.14 | 70,892 |
| Portugal | 7,928 | 3,868 | 48.8% | 62.3% | 3.51 | 131,232 |
| Slovenia | 8,156 | 3,931 | 48.2% | 72.4% | 2.32 | 39,062 |
| Lithuania | 2,073 | 805 | 38.8% | 64.3% | 2.66 | 45,958 |
| Spain | 25,732 | 9,768 | 38.0% | 55.1% | 4.39 | 218,607 |
| Italy | 7,113 | 2,571 | 36.1% | 56.2% | 3.85 | 356,436 |
| Malta | 330 | 111 | 33.6% | 49.4% | 3.93 | 290,052 |
| Estonia | 4,634 | 1,471 | 31.7% | 57.4% | 3.49 | 42,000 |
| Slovakia | 4,124 | 1,213 | 29.4% | 58.7% | 2.72 | 45,884 |
| Croatia | 20,263 | 5,633 | 27.8% | 56.2% | 2.63 | 37,016 |
| Netherlands | 4,350 | 1,139 | 26.2% | 46.0% | 3.62 | 503,700 |
| Latvia | 5,688 | 1,439 | 25.3% | 46.7% | 3.88 | 38,470 |
| Austria | 3,270 | 820 | 25.1% | 41.1% | 3.93 | 511,340 |
| Belgium | 5,030 | 1,223 | 24.3% | 48.5% | 3.48 | 270,000 |
| Luxembourg | 412 | 95 | 23.1% | 41.0% | 3.82 | 414,993 |
| Germany | 38,923 | 8,005 | 20.6% | 37.3% | 5.02 | 251,868 |
| Ireland | 1,800 | 371 | 20.6% | 35.8% | 4.79 | 240,000 |
| France | 23,929 | 4,443 | 18.6% | 37.5% | 6.22 | 245,000 |
| Finland | 3,494 | 588 | 16.8% | 32.3% | 4.81 | 242,922 |
Outside the euro area, with lot counts, the same measure runs: Poland 56.5 per cent on 81,277 lots, Bulgaria 45.0 per cent on 10,717, Czechia 43.6 per cent on 20,396, Hungary 37.5 per cent on 6,504, Romania 33.7 per cent on 212,953, Sweden 23.8 per cent on 7,220 and Denmark 22.2 per cent on 2,329. Norway is at 17.6 per cent on 1,254 lots and Switzerland 12.9 per cent on 2,667.
Two patterns in that table sit outside the rankings. Croatia and Slovakia both sit near 28 per cent on single bids, but they also have high shares of lots at two bids or fewer, 56.2 and 58.7 per cent, which places them close to Spain and Italy despite very different average lot values. Mean bids spread wider still. France averages 6.22 bids a lot and Finland 4.81, against 2.07 for Cyprus and 2.32 for Slovenia. Some of that spread is market structure and some is how buyers count and publish bids, and the bulk data cannot separate the two.
Zero bids and one bid, by country
A lot that receives no bid is a failed tender, and the packages record it. Of the 584,516 lots carrying a bid count, 65,339 record zero, or 11.2 per cent. Add the zero-bid lots to the single-bid lots and you get the share of lots that reached, at most, one firm.
| Country | Lots with a bid count | Zero bids | Zero % | One bid | Zero or one % |
|---|---|---|---|---|---|
| Poland | 81,902 | 594 | 0.7% | 45,944 | 56.8% |
| Greece | 5,339 | 58 | 1.1% | 2,835 | 54.2% |
| Slovenia | 8,991 | 833 | 9.3% | 3,931 | 53.0% |
| Bulgaria | 12,203 | 1,486 | 12.2% | 4,828 | 51.7% |
| Lithuania | 2,611 | 538 | 20.6% | 805 | 51.4% |
| Portugal | 8,052 | 119 | 1.5% | 3,868 | 49.5% |
| Romania | 262,404 | 49,451 | 18.8% | 71,749 | 46.2% |
| Czechia | 21,248 | 843 | 4.0% | 8,884 | 45.8% |
| Italy | 8,354 | 1,241 | 14.9% | 2,571 | 45.6% |
| Estonia | 5,496 | 862 | 15.7% | 1,471 | 42.4% |
| Spain | 26,917 | 1,177 | 4.4% | 9,768 | 40.7% |
| France | 26,249 | 2,266 | 8.6% | 4,443 | 25.6% |
| Germany | 40,726 | 1,754 | 4.3% | 8,005 | 24.0% |
| Ireland | 1,873 | 72 | 3.8% | 371 | 23.7% |
| Switzerland | 2,874 | 207 | 7.2% | 344 | 19.2% |
The zero column and the one column do not track each other. Poland tops the single-bid table with almost no failed lots, 0.7 per cent. Romania fails 49,451 lots, 18.8 per cent of its own total, and still lands at 33.7 per cent single-bid. Lithuania fails one lot in five and Estonia one in six. A zero-bid total is part market outcome and part record-keeping, because a cancelled or retendered procedure can leave its original lot in the file with a zero.
Countries differ in where the failure sits, at the first attempt or the second. A bidder spending a week of estimating time should read the two columns together, and a buyer looking for a benchmark should compare itself against its own country on both.
Bigger lots do not draw more bids
If competition were a function of money, the dearest decile of lots would attract the most bidders. It does not. Sort the 122,032 euro-area lots with a usable euro value by value and cut them into ten equal groups of about 12,203 lots each. The single-bid rate barely shifts from decile to decile.
| Decile | Value range, EUR | Lots | Mean lot, EUR | One bid % |
|---|---|---|---|---|
| 1 (cheapest) | 1,000 to 7,629 | 12,204 | 3,719 | 31.3% |
| 2 | 7,630 to 22,846 | 12,204 | 14,379 | 33.4% |
| 3 | 22,848 to 48,488 | 12,203 | 34,553 | 31.3% |
| 4 | 48,496 to 91,720 | 12,203 | 68,220 | 30.3% |
| 5 | 91,724 to 160,864 | 12,203 | 124,108 | 29.0% |
| 6 | 160,870 to 260,151 | 12,203 | 209,179 | 31.1% |
| 7 | 260,152 to 417,112 | 12,203 | 331,974 | 33.2% |
| 8 | 417,119 to 760,500 | 12,203 | 563,723 | 33.0% |
| 9 | 760,782 to 1,976,100 | 12,203 | 1,204,605 | 29.9% |
| 10 (dearest) | 1,976,321 to 500,000,000 | 12,203 | 13,680,162 | 27.0% |
The cheapest decile runs at 31.3 per cent single-bid and the dearest at 27.0 per cent. The same result holds in bands: lots between EUR 1,000 and EUR 10,000 sit at 31.9 per cent single-bid, lots between EUR 50,000 and EUR 250,000 at 30.0 per cent, lots between EUR 250,000 and EUR 1m at 32.7 per cent, lots between EUR 1m and EUR 10m at 28.2 per cent and lots above EUR 10m at 27.1 per cent. Split at the EU advertising thresholds, lots under EUR 200k run at 30.9 per cent, lots between EUR 200k and EUR 1m at 32.6 per cent and lots at EUR 1m or more at 28.0 per cent.
Money does not decide whether a lot is contested, at least in this six-month window. The weak downward slope at the top is real and small, and two mechanical explanations compete with any behavioural one. Larger lots take longer to bid and need more capacity, which may thin the field, and larger lots are also the ones most likely to be renewed under an existing framework or published as a call-off, where the competition happened earlier. The data carries no framework flag, so that cannot be tested. A bid manager who expects a EUR 20m lot to be crowded is working against the pattern in this file.
Value is concentrated more sharply than bids. The dearest one per cent of priceable euro-area lots, 1,220 lots, carry 51.0 per cent of all the priceable value in the cut, and the dearest ten per cent carry 84.3 per cent.
Where the money is, and how much of it went uncontested
The priceable cut holds 122,032 euro-area lots worth EUR 198.1bn in stated euro value, with no conversion applied. The median lot is EUR 160,856, the 90th percentile EUR 1,976,100, which leaves nine lots in ten below EUR 2.0m. Of that total, EUR 61.2bn across 37,761 lots went on a single bid, which is 30.9 per cent of all priceable euro-area value.
Country totals move the picture in a way country rates do not. Rank the countries by the share of priceable value that went on one bid and Lithuania leads at 74.1 per cent of its EUR 1.98bn, then Greece at 63.9 per cent of EUR 6.00bn, Slovenia at 57.5 per cent of EUR 4.55bn and Portugal at 51.1 per cent of EUR 13.66bn. Ireland sits at the other end, 14.5 per cent of EUR 3.55bn.
One number is used loosely elsewhere. The published study carries EUR 61.2bn on single-bid lots, 30.9 per cent of priceable value. Both reproduced exactly from this database, to the euro, along with the EUR 198.1bn total, the EUR 160,856 median and the EUR 1,976,100 90th percentile. The same filters give the same total six months later.
Sectors: thin markets and deep markets
CPV divisions group the notices by what is bought. The euro-area table below takes the divisions with 500 lots or more, sorts them by single-bid rate, and puts the median lot value beside each one.
| CPV division | What it covers | Lots | One bid % | Median lot, EUR |
|---|---|---|---|---|
| 38 | Laboratory, optical and precision equipment | 3,390 | 58.3% | 162,490 |
| 48 | Software packages and information systems | 2,505 | 53.1% | 340,196 |
| 50 | Repair and maintenance services | 7,080 | 44.2% | 210,000 |
| 35 | Security, fire-fighting, police and defence equipment | 1,359 | 42.5% | 250,000 |
| 64 | Postal and telecommunications services | 955 | 41.6% | 346,136 |
| 24 | Chemical products | 1,028 | 41.1% | 86,416 |
| 42 | Industrial machinery | 1,581 | 40.9% | 248,751 |
| 34 | Transport equipment and auxiliary products | 7,522 | 39.2% | 219,950 |
| 72 | IT services | 5,536 | 38.2% | 450,000 |
| 80 | Education and training services | 3,851 | 38.2% | 133,253 |
| 33 | Medical equipment, pharmaceuticals and personal care | 40,178 | 35.7% | 46,000 |
| 85 | Health and social work services | 3,534 | 34.2% | 389,000 |
| 90 | Sewage, refuse, cleaning and environmental services | 9,128 | 21.0% | 213,727 |
| 79 | Business services | 6,696 | 19.9% | 237,329 |
| 77 | Agricultural, forestry and horticultural services | 4,545 | 18.4% | 100,808 |
| 71 | Architectural and engineering services | 12,301 | 16.3% | 214,210 |
| 45 | Construction work | 20,873 | 11.5% | 400,000 |
At division level the thin markets sharpen. Vehicle repair and maintenance, division 504, comes in at 69.7 per cent single-bid on 1,003 lots. Measuring and testing instruments, 384, at 58.1 per cent on 954 lots. Laboratory services, division 380, at 55.0 per cent on 1,303 lots. Software packages, division 480, at 52.7 per cent on 1,264 lots. Repair of vehicles and related equipment, 501, at 50.5 per cent on 1,210 lots. Software consultancy, 722, at 47.4 per cent on 2,191 lots. Office and computing machinery, including medical consumables coded under 330, at 42.2 per cent on 2,209 lots.
The deep end of the table is mostly construction, plant and services, the categories where a local supplier list already exists. Depth shows in a second column too, the share of lots in a division that drew five or more bids. Construction work leads there at 46.7 per cent, business services at 44.9 and agricultural services at 43.9, against 23.3 per cent for research services and 25.2 per cent for IT services.
The pattern points at categories where one authorised distributor, service agent or incumbent holds the practical route to the buyer. It is a hypothesis and not a measurement. CPV divisions are not markets, a division can contain several unrelated markets, and a single-bid rate in a division says nothing about which sub-category drove the number. That would need the lot-level CPV code and a supplier register, and neither is in this build.
Buyer type: hospitals see fewer bidders than councils
The notices carry no buyer type field, so we classified buying organisations by keywords in the buyer name. Each lot falls into one class, and the rest sit in an "other" bucket.
| Buyer class | Lots | One bid % |
|---|---|---|
| Hospital and health bodies | 9,680 | 43.6% |
| Universities and schools | 6,512 | 35.8% |
| National agencies and ministries | 12,216 | 35.8% |
| Utilities and transport operators | 4,925 | 27.0% |
| Local and regional government | 10,562 | 24.9% |
| Other and unmatched names | 129,083 | 27.6% |
Health bodies post 43.6 per cent single-bid, the highest of the named classes and far above local government at 24.9 per cent. That matches the sector table, where medical goods and laboratory equipment have the thinnest fields, and it matches the buyer list in the next section, which hospitals populate. The classification is a keyword rule on a name string, so the ordering is indicative and not exact, and the 129,083 lots in the unmatched bucket include many bodies a proper entity register would classify. Read the gap between health and local government as a signal and the exact percentages as estimates.
Buyers that never see a second bid
Grouping lots by buying organisation exposes a population the country tables hide: buyers who award repeatedly and almost never see more than one bid. Buyers with at least five lots and a single-bid rate of 80 per cent or more number 866 when grouped by buyer name and country, and 865 when grouped by name alone. The difference is one body whose name appears under two country codes. Buyers with ten lots or more and a 100 per cent single-bid record number 113.
| Buyer | Country | Lots | One bid % |
|---|---|---|---|
| U.A.T. Judetul Constanta, Consiliul Judetean | Romania | 884 | 83.5% |
| Szpital Uniwersytecki nr 1 im. dr. A. Jurasza, Bydgoszcz | Poland | 595 | 83.5% |
| Szpital Wolski | Poland | 405 | 89.6% |
| Unidade Local de Saude Santa Maria | Portugal | 337 | 97.9% |
| Directia Generala de Asistenta Sociala si Protectia Copilului Prahova | Romania | 305 | 100.0% |
| Warminsko-Mazurskie Centrum Chorob Pluc, Olsztyn | Poland | 238 | 98.7% |
| Unidade Local de Saude de Braga | Portugal | 141 | 100.0% |
| Nemzeti Kommunikacios Hivatal | Hungary | 136 | 100.0% |
| Cámara Oficial de Comercio de Barcelona | Spain | 143 | 86.7% |
Almost every named buyer is a hospital or a regional health and social care body, and most sit in Romania, Poland, Portugal, Greece and Hungary. There is a mechanical component here. A hospital that buys consumables month after month in small lots, from a supplier holding the distribution agreement for the product, will record one bid repeatedly. Nothing here suggests wrongdoing. The count of bids is a weak contest measure in exactly the categories where the money is routine.
Buyer size and competition are related only at the top. Buyers with a single lot in the window average 31.0 per cent single-bid, buyers with 2 to 3 lots 30.8 per cent, buyers with 4 to 10 lots 30.0 per cent, buyers with 11 to 50 lots 31.5 per cent, and buyers with 51 lots or more 39.8 per cent. Concentration is high: the top 1 per cent of buyers by lot count, 265 organisations, hold 49.1 per cent of all lots in the cut, and the top 10 per cent hold 79.8 per cent.
Repeat winners and the shape of the supplier tail
Winners appear as supplier name strings, lower-cased and stripped of punctuation. The winners table holds 725,856 rows, of which 16,924 carry no usable name. After normalising and counting distinct award notices, there are 135,814 distinct supplier names across 368,223 notice-supplier pairs.
| Supplier track record | Suppliers | Share of suppliers | Notices won | Share of notice-supplier pairs |
|---|---|---|---|---|
| One notice | 90,950 | 67.0% | 90,950 | 24.7% |
| Two or more | 44,864 | 33.0% | 277,273 | 75.3% |
| Five or more | 11,604 | 8.5% | 194,215 | 52.7% |
| Twenty or more | 1,840 | 1.4% | 113,292 | 30.8% |
| One hundred or more | 240 | 0.2% | 53,589 | 14.6% |
Two thirds of the supplier names turn up once. At the far end, 240 names take 14.6 per cent of all notice-supplier pairs, and 1,840 take 30.8 per cent. Ranked, the top 1 per cent of suppliers by notice count, 1,358 names, hold 27.9 per cent of pairs; the top 5 per cent hold 45.4 per cent and the top 10 per cent hold 54.9 per cent. The published study reports 29.8 per cent for its top 1 per cent share on its own build. Our re-run gives 27.9 per cent on this database. The difference is the denominator and the way winner names are collapsed, not a different window.
| Supplier name | Notices won | Country pattern |
|---|---|---|
| farmexim s a | 1,141 | Romania, pharmaceuticals distribution |
| b braun medical | 1,129 | medical devices across several countries |
| urtica | 1,057 | Romania, pharmaceuticals |
| fresenius kabi romania | 956 | Romania, hospital pharmaceuticals |
| bialmed | 937 | Poland, pharmaceuticals |
| dr max | 928 | Romania, pharmacy chain |
| farmaceutica remedia distribution | 725 | Romania |
| asclepios s a | 693 | Romania |
| alliance healthcare romania | 692 | Romania |
| strabag | 687 | construction across central Europe |
| eurovia cz | 685 | Czechia, road construction |
| farmacol logistyka | 675 | Poland |
The list is dominated by pharmaceutical wholesalers and distributors, with construction groups behind them, the same market the hospital buyers above operate in, seen from the other side. Pair the two and you get relationships: 5,466 buyer-supplier pairs with five or more awards in six months, and 336 pairs with twenty-five or more. The heaviest single pairing is Klinički bolnički centar Zagreb with medika d d at 202 notices, then oktal pharma and phoenix farmacija with the same hospital at 199 and 184.
A repeat winner is not proof of a competition failure. Framework call-offs, medical consumables supplied under distribution agreements, and utilities with few qualified contractors all produce repeat winners by design. What the numbers establish is that the EU award data is a long series of small, repeated, largely uncontested transactions in the health and construction supply chains, and that the buyers and suppliers concerned can be named.
Who wins when there is only one bid
Single-bid lots with a clear single supplier can be traced through the notice that names it. Of the EUR 61.2bn on single-bid euro-area lots, EUR 42.5bn, or 69.4 per cent, sits in notices naming exactly one supplier, across 19,775 lots. Split that attributed value by the supplier's overall notice count and most of it sits with occasional winners, not dominant ones.
| Supplier track record | Single-bid lots | Value, EUR | Share of all single-bid value |
|---|---|---|---|
| One to four notices | 11,752 | 28,562,402,945 | 46.7% |
| Five to nineteen notices | 5,290 | 9,276,348,122 | 15.2% |
| Twenty to ninety-nine notices | 2,402 | 4,085,772,169 | 6.7% |
| One hundred notices or more | 331 | 569,933,437 | 0.9% |
The single-bid money does not sit with the serial winners identified above. Nearly half goes to suppliers who appear in one to four award notices in the window. Uncontested procurement is mostly small and occasional suppliers taking the only bid on a small and occasional lot, and the serial distributors in the previous section account for under one per cent of the attributed single-bid value.
One caveat matters: 14,540 single-bid lots have no winner recorded at all, out of 28,565 lots in the cut with no winner row. Any statement about who wins single-bid work rests on the roughly 70 per cent of the value where the supplier is named.
Frameworks and lot structure
The bulk data carries no framework flag, so we fell back on a proxy: notices with the word "framework" in the title. Those notices hold 5,169 lots at 15.9 per cent single-bid, against 513,835 lots in other notices at 36.3 per cent. Framework arrangements compete before the contract is placed, and the award notices that follow record the call-off, which is where 15.9 and 36.3 part company. The proxy is crude, because a notice titled "framework agreement" can still be a competitive re-opening and a call-off can be published under its own name.
Split notices by lot count and no strong relationship appears between lot structure and competition on the lot. Notices with one lot run 30.7 per cent single-bid, notices with 2 to 3 lots 27.8 per cent, 4 to 10 lots 29.4 per cent, 11 to 50 lots 29.3 per cent, and notices with 51 lots or more 25.0 per cent. The average award notice carries 2.06 lots, and the largest single notice carries 373.
Two ways to count the same tender
Lot-level and notice-level reporting measure different things, and on the same data they give different answers. Across the 83,956 award notices that contain at least one lot with a bid count, 23,301 notices, or 27.8 per cent, have every lot at a single bid. A further 28,474 notices, 33.9 per cent, contain at least one single-bid lot. Counted on the lot, all countries together give 36.1 per cent. Three defensible headlines, 27.8, 33.9 and 36.1, all from one six-month file.
Lot splitting explains the gap. Bidders bid per lot, not per notice, and the bid count at notice level comes from a different field than the bid count at lot level. Where both are present they often disagree: 11,637 lots sit in notices recording seven bids and on lots recording one, 11,134 lots sit in notices recording five with one on the lot, and 74,788 lots sit in notices and lots that both record one. A notice-level average flatters a lot nobody wanted, because a popular sibling lot can carry the average.
Buyers run into the same translation problem. On the notice grain, 221,931 award notices came from 36,038 buying organisations, which gives 6.2 awards per buyer; that matches the published study. On the lot grain, the same 519,004 lots came from 26,490 buyers, which gives 19.6 lots per buyer. On the lot grain, 9,274 buyers never award more than one lot in six months. One figure without its grain is how two correct analyses end up disagreeing in public.
The fields that would explain this are almost always empty
The way to test whether open procedures compete better than negotiated ones is to read the procedure field. The field is filled on 4,023 of 221,931 award notices, 1.8 per cent. The award criterion field is filled on 4,306 notices, 1.9 per cent. Where they are filled the values are: open procedure 3,001 notices, negotiated 566, restricted 456; lowest price 1,112, most economically advantageous tender 3,165, mixed 29, other 8.
With 1.8 per cent coverage, nothing about procedure design and competition can be read from this data, and any published claim that tries is reading a rounding error. The SME bidder counts are populated on a small minority of notices, and any framework or call-off flag is absent. A buyer that wants its own performance compared has to publish those fields in the notice, and the practical version of that request is one line in a tender template.
Ireland read against the rest
Ireland has 1,800 lots in the cut, 371 of them single-bid, 20.6 per cent, against 29.2 per cent in the euro area and 36.1 per cent across all countries. Irish lots at two bids or fewer are 35.8 per cent, on a mean of 4.79 bids and a median lot value of EUR 240,000. Of 1,276 Irish lots with a usable euro value, the priceable total is EUR 3.55bn, and EUR 515.7m of it, 14.5 per cent, went on single-bid lots. That is the second-lowest single-bid value share among the seventeen euro-area countries in this build, behind Latvia at 12.9 per cent and below Finland at 18.5 per cent, Germany and France at 19.5 per cent.
The sector pattern inside Ireland is the familiar one. Laboratory and optical equipment, CPV 38, runs 47.4 per cent single-bid on 76 lots, construction work, CPV 45, 12.1 per cent on 257 lots. Architectural and engineering services sit at 17.6 per cent on 239 lots, business services at 12.1 per cent on 174 lots. The largest Irish buyers by lot count are the Health Service Executive with 104 lots and a 23.1 per cent single-bid rate, Dublin City Council with 68 lots at 35.3 per cent, Iarnród Éireann with 67 lots at 14.9 per cent, Trinity College Dublin with 49 lots at 44.9 per cent, and the Department of Housing, Local Government and Heritage with 47 lots at 19.1 per cent. The Irish detail sits in a separate build, published as the Ireland public contracts study, because the Irish lot counts are small enough that a single buyer can move a national percentage.
Two cautions apply to the Irish figures. Irish priceable coverage is 70.9 per cent, so EUR 3.55bn covers 1,276 of 1,800 lots and the value share could move with better reporting. Irish lot splitting is modest by EU standards, which raises the lot-level single-bid rate less than it does in Romania or Poland. Irish buyers split awards into fewer, larger lots than their continental counterparts, which gives more bidders a reason to bid on each one.
What this means for you
If you bid into EU public contracts, the numbers point to two different plays. The thin categories, laboratory and optical equipment at 58.3 per cent single-bid, software packages at 53.1 per cent, vehicle repair at 69.7 per cent, are thin because buyers accept one bid, and there is money in them. The median lot in the thin categories is between EUR 160,000 and EUR 450,000, enough to cover a bid preparation. Many of those lots sit inside distribution agreements, so the work is to become the second bidder on a contract already served, and to watch the renewal date and not the notice date.
If you are a buying organisation, the useful comparison is against your own country and your own buyer class, and not against the euro-area average. Health bodies run 43.6 per cent single-bid and local government 24.9 per cent, so a hospital at 30 per cent is doing better than its peer group and a council at 30 per cent is doing worse. Two internal measures are worth a quarterly count: the share of your lots with one bid, and the share of your priceable value awarded on those lots. Both come out of a public notice file in an afternoon, with the queries in the next section.
If you are a journalist or an analyst, the first question about any competition figure is which grain it uses, because 27.8, 33.9 and 36.1 per cent are all correct statements about this window. The second is the denominator, because a rate quoted without a lot count can come from a country with 330 lots or a country with 81,277. The third is the currency and the value filter, because 30.9 per cent of value and 62 per cent of value usually differ by a filter and not a finding.
If you set procurement policy, the most useful change visible in this data is not a new threshold or a new procedure. It is the publication of the fields that are empty. Procedure type is filled on 1.8 per cent of award notices and the award criterion on 1.9 per cent, so the two variables that would let anyone test whether an open procedure attracts more bidders than a negotiated one are missing from 98 of every 100 notices. Directive 2014/24/EU already provides for the data, and the delivery gap sits in the notice templates and the national portals that feed them. The European Commission public procurement scoreboard and the Commission procurement pages track national performance on a different basis and sit well beside these lot counts.
What we do not know
Country reporting differences are the largest single source of uncertainty here. Countries split awards into lots at very different rates, and the lot counts show it: Romania contributes 212,953 lots, Poland 81,277, Germany 38,923, while France contributes 23,929 on a far bigger economy. A country that publishes one notice per product line and one that publishes one notice per contract can run identical procurement and land on different single-bid rates on our measure. We cannot correct for that, because the data carries no field for how the buyer chose to split.
Value reporting is worse than bid reporting and it varies by country. Of the euro-area lots, 122,032 of 172,978, or 70.5 per cent, carry a usable euro value between EUR 1,000 and EUR 500m. The spread by country is wide: Switzerland 98.2 per cent coverage, Italy 95.4 per cent, Spain 88.4 per cent, France 81.3 per cent, Ireland 70.9 per cent, Germany 39.1 per cent and Sweden 33.0 per cent. Any statement about German or Swedish procurement value in this window describes the third of lots that report a value and not the whole market. Outside the euro area, raw values are unusable for totals, since they are recorded in the notice currency without conversion. The average recorded value on a Polish lot in this file is above PLN 800m, a data artefact and not a market fact. Every euro figure in this article is euro-area only, and there is no EU-wide value total anywhere in it.
Bid counts are missing across a large part of the file. Of 761,062 lot rows in the notice tables, 584,516 carry a bid count; 65,339 record zero, 73 claim more than 200, and 519,004 remain in the main cut. A missing bid count could mean the buyer did not publish it, the field was dropped at the national portal, or the lot was cancelled before award. The zero-bid lots have the same problem at a smaller scale, because a cancelled procedure can leave a lot with a zero, and 14,540 single-bid lots have no winner recorded at all.
The window is six months of publication dates in 2026, March to August. That is long enough to measure a rate, too short to measure a change. The monthly packages carry a load stamp in the year field, not a publication year, which is why this article makes no trend claim and the published study does not either. Award notices published in the window can relate to tenders advertised in 2024 or earlier, so the file is not a picture of six months of tendering activity. Below-threshold procurement appears only where a buyer chose to advertise it, which biases the file toward larger contracts, and voluntary ex ante transparency notices are awards with a single supplier by construction.
Three more limits are structural. There is no framework or call-off flag in the data, so the framework analysis uses a title keyword. The data carries no contract-level value, duration or payment record, so nothing here measures overspend. Buyer and supplier identities are name strings: 866 buyers sit at 80 per cent or more single-bid when grouped by name and country and 865 when grouped by name alone, and the 135,814 supplier names include the same company spelled several ways, which is why the top 1 per cent supplier share of 27.9 per cent is a range and not a point.
A note on the published study. This article re-uses the same database and did not repeat the extraction from the TED packages. Where we re-ran a published figure we say so, and where we could not reproduce it exactly we say that too. The published EUR 61.2bn single-bid value, the 30.9 per cent share, the EUR 198.1bn total, the EUR 160,856 median, the EUR 1,976,100 90th percentile, the 519,004 lot count, the 29.2 per cent euro-area single-bid rate and the 6.2 awards per buyer all reproduced exactly and come from our own re-run. Two published numbers did not reproduce exactly on this database. The top 1 per cent supplier share is 29.8 per cent in the study and 27.9 per cent in our re-run, on a different denominator. The distinct supplier count is 135,815 in the study and 135,814 in our re-run. Both gaps are normalisation or denominator effects. The thin-buyer count of 866 and the 113 never-contested buyers reproduced exactly, and so did the country and sector rates quoted above. Two further published figures, the 44.4 per cent of buyers with a single award and the 21.6 per cent of award notices with more than one winner, are counted on the notice grain and we did not recompute them here; our lot-grain equivalent of the first is 35.0 per cent, from 9,274 buyers holding exactly one lot out of 26,490 buyers in the cut.
How to reproduce this
The source behind all of this is free and public. Two routes reach it and they are not equivalent, so the choice matters.
The first route is the TED open data portal, which serves the eForms packages this analysis uses and requires a registered key for the API and for bulk download. Begin with the TED developer documentation, register for a key, then pull the monthly award packages into SQLite. The notice search interface at ted.europa.eu is where you check any single notice or contract by hand, and the CPV code meanings come from the EU vocabulary for public procurement. One warning for readers following the published study: the TED packages page cited there, ted.europa.eu/en/monitor/packages, returned HTTP 404 when we checked it on 26 September 2026, and the developer portal is the current route to those files.
The second route is the free CSV subset on data.europa.eu, covering contract award notices up to 2023 with no key needed. It is the fastest way to reproduce the country and competition patterns for the eForms era before our window. The download was checked from a shell on 26 September 2026 and returns a zip containing Export_OpenDataCAN_year2018.csv style files:
curl -L -o ted-award-notices-2018-2023.zip \
"https://data.europa.eu/api/hub/store/data/ted-contract-award-notices-2018-2023.zip"
unzip -o ted-award-notices-2018-2023.zipThe dataset landing page is the TED CSV dataset on data.europa.eu, and the series stops at 2023, which is why the six months analysed here come from the packages route.
Every figure in this article comes from two tables in a single SQLite database. The main competition cut is:
select count(*) lots, sum(case when nt=1 then 1 else 0 end) single_bid
from lotx_clean
where nt between 1 and 200
and cc in ('AUT','BEL','HRV','CYP','EST','FIN','FRA','DEU','GRC','IRL',
'ITA','LVA','LTU','LUX','MLT','NLD','PRT','SVK','SVN','ESP');The value cut adds the priceable filters and nothing else:
select count(*), round(sum(val)), sum(case when nt=1 then val end)
from lotx_clean
where nt between 1 and 200 and val between 1000 and 5e8
and cc in ('AUT','BEL','HRV','CYP','EST','FIN','FRA','DEU','GRC','IRL',
'ITA','LVA','LTU','LUX','MLT','NLD','PRT','SVK','SVN','ESP');And the buyer cut groups the same table by the buyer string:
select buyer, cc, count(*) lots,
round(100.0*sum(case when nt=1 then 1 else 0 end)/count(*),1) one_bid_pct
from lotx_clean group by buyer, cc
having count(*) >= 5 and one_bid_pct >= 80 order by lots desc;Outputs differ from the published study only where the grouping rule differs, as noted above. Our full set of 55 cuts, each with the value, the SQL that produced it and the column it came from, is written to figures-eu.json alongside this draft, so any number above can be re-run. Questions, corrections and requests for the underlying CSV extracts go through the what we do page, and the short version for media is the press page for the procurement study. Other data studies sit under Rennwick insights.
Sources
- TED, the Tenders Electronic Daily monthly eForms packages plus the developer documentation, European Publications Office. https://developer.ted.europa.eu/docs/ (checked 26 September 2026). Source of the March to August 2026 notice data.
- TED notice search and the notice viewer. https://ted.europa.eu/en/search/result (checked 26 September 2026). Used to spot-check individual notices and awards.
- TED CSV subset, contract award notices and contract notices, from the EU open data portal. https://data.europa.eu/data/datasets/ted-csv?locale=en (checked 26 September 2026; series ends 2023). Used as the reproduction route.
- Directive 2014/24/EU, the public procurement directive. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32014L0024. Used for procedural and reporting context.
- Directive (EU) 2019/1024 on open data and re-use of public sector information. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32019L1024. Used for context on open data.
- European Commission page on public procurement in the single market. https://ec.europa.eu/growth/single-market/public-procurement_en (checked 26 September 2026). Used for context on policy.
- European Commission Single Market Scoreboard, the public procurement chapter. https://single-market-scoreboard.ec.europa.eu/business-framework-conditions/public-procurement_en (checked 26 September 2026). Used as the official comparator on competition indicators.
- Eurostat data on government expenditure by function. https://ec.europa.eu/eurostat/databrowser/view/gov_10a_exp/default/table (checked 26 September 2026). Used only for the size of public purchasing, and for no figure in this article.
- EU vocabulary, the Common Procurement Vocabulary. https://single-market-economy.ec.europa.eu/single-market/public-procurement/digital-procurement/common-procurement-vocabulary_en (checked 26 September 2026). Used to label the CPV divisions in the sector table.
- Rennwick, the European public procurement competition study, published March to August 2026 build. https://www.rennwick.com/european-public-procurement-competition-study/. Source of the headline figures re-run in this article.
Revision log
Version 1.0, dated 26 September 2026. First publication of the deeper-cut piece. Data window March to August 2026, TED monthly eForms packages, loaded on 14 September 2026. 55 cuts recomputed on 26 September 2026 from the same SQLite build. Method change from the published study: none in the filters, which are identical, plus two additions, the lot-value decile table at full precision and the notice-to-lot translation cut. Corrections will be logged here with a date and the previous value.