What bid rigging looks like in data
Bid rigging means that bidders agree in advance who wins and at what price. From outside, the procedure looks fair: there are several bids and the price falls. But the agreement leaves traces in the data: in prices, in the order and timing of bids, in who takes part and in who wins year after year.
The Open Contracting Partnership has collected 73 red flags for all stages of procurement, and a separate group of them describes collusion risk. Each flag is calculated from procurement fields: the bid price, the filing date, the winner and the bidder’s contacts.
One flag means almost nothing. Identical prices happen in a market with a common manufacturer price list, and a constant winner can be a supplier that is simply cheaper. A signal for review appears when several signs match at once for one group of bidders: prices, links and the history of wins.
A hypothetical example: Supplier A and Supplier B have bid together in Buyer C’s tenders for three years. They win in turns, the gap between their prices is less than one percent, and their bids are filed a minute apart. Each sign alone can be explained, but together they give the procedure a high risk score, and it goes to a reviewer.
The data you need: tender notices and protocols, all bids with prices (not only the winning one), filing date and time, bidder details with contacts and owners, and the buyer’s contract history over several years. Without the losing bids, you cannot calculate most signs of collusion.
Signs in bids and prices
In the Open Contracting Partnership guide, these signs in bids point to collusion risk:
1. Identical prices from different bidders. This happens when bidders agreed on a price or when the buyer disclosed competitors’ prices.
2. The second-place price is too close to the winner’s. A bidder fakes competition and does not stop a partner from winning.
3. Too wide a spread of prices in one procedure. Losing bids agreed in advance are priced far above the winning one.
4. Bids are filed in the same order at different stages of the tender.
5. Winner rotation: bidders win in turns, and each gets a contract at an inflated price.
6. Pairs of bidders who enter several tenders together, with the same one always winning.
7. Similar documents from different bidders: the same wording and layout, as if one person prepared all the bids.
The Korean BRIAS system looks for similar patterns: bids filed at the same time; one company steadily cheaper while the others are clearly more expensive; unusual price swings over time.
What AI does: for each procedure, it compares prices with the market distribution for the same product, looks for repeating pairs and chains of wins over the years, and gives the procedure a risk score. The reviewer gets a list of tenders sorted by score and sees which signs fired in each case.
Shared addresses, phones and IPs
The second group of signs concerns the bidders themselves, not prices. If two competitors share a phone, address, contact person or email, they may be one company under different names. The Open Contracting Partnership suggests flagging both the procedure and both bidders in such a case.
A separate sign: a bidder’s contacts match the buyer’s contacts. According to OCP, this is sometimes what a shell company looks like, or a scheme where the winner prepares the bids of non-existent competitors.
Colombia showed the scale of such matches. The Océano platform of the Comptroller General’s Office analyzed 8,678,464 public contracts from 2014–2019 and found 8,749 companies sharing one phone number and 54,155 cases where one representative stood behind several companies.
The IP address and filing time of a bid are stored by the e-procurement platform if it keeps a technical log. For AI, this is one more field to compare: two bidders who filed from the same address get a shared link in the case graph.
Matching contacts are not a verdict either: companies of one group may share an accountant or a business center. How AI joins registers and shows hidden links is covered on the page about affiliation.
The FAS experience: from Big Digital Cat to Anti-Cartel
Russia’s Federal Antimonopoly Service (FAS) has been building automatic cartel detection for several years. The first module of its Big Digital Cat system was launched in autumn 2019.
The state Anti-Cartel information system has worked since 1 August 2025. According to FAS, it analyzes data from Russia’s Unified Procurement Information System and the state Torgi platform every day, with elements of machine learning, and covers 100% of procurement procedures.
For a buyer and an audit body, this means procurement data already goes through automatic checks at the regulator. Your own check is useful earlier: before the results are final, while the contract is not signed and the money has not left.
The system does not decide that a cartel exists. Staff of the authority review the signal, and a commission of the competition authority or a court establishes the violation.
How AI helps audit bodies read documents and match payments is described on the page about AI for control and audit.
Korea: BRIAS and 40,000 procedures a year
The Korea Fair Trade Commission (KFTC) has run the BRIAS system since 2006. The system collects data on public tenders above thresholds, for example from KRW 100 million for goods and services, and calculates statistics on them. In 2023 it ran 40,170 analyses, and in 2018, 25,520.
Since December 2023, more than 700 organizations, including local public companies, must submit tender data. The KFTC compares BRIAS to a surveillance camera that sees even narrow alleys: the system does not punish, it shows patterns that look like collusion.
The KFTC’s main caveat: the system does not prove collusion, but it gives investigators important leads. That is how AI should work in control: a flag means risk, and the decision belongs to the reviewer.
An example from the same presentation: supplies of coagulants for water treatment in 2019–2023. Two companies split volumes and prices in advance and won 14 and 17 contracts worth KRW 438 billion, 43% of the market.
The KFTC used BRIAS along the whole case, from the first suspicions to confirming the collusion. The commission, not the system, made the decision and set a fine of KRW 391 million. More on the Korean experience on the page about the BRIAS system.
What a buyer should do
A buyer sees its own tenders before anyone else and can check them before signing a contract. The order is as follows.
1. Export all bids for your procedures over two or three years, with the prices of the losing bidders.
2. Calculate simple signs: a single bidder, identical prices, a small gap between first and second place, repeating pairs of bidders.
3. Compare bidders’ contacts with each other and with your own: phones, addresses, email, contact persons.
4. Mark the procedures where two or more signs fired, and pass them to internal control together with the export.
5. If the documents confirm the signs, management decides whether to send the materials to the competition authority.
A flag alone is no reason to cancel a tender or accuse a bidder. It is a signal for review, and the conclusion about a violation is made by the reviewer, the competition authority or a court. How to combine such indicators into one system is described on the page about red flags in procurement.