In June 2025, the OECD updated its Guidelines for Fighting Bid Rigging in Public Procurement. The document includes a detection checklist — a list of patterns that should raise red flags for anyone reviewing procurement data.
Reading it is uncomfortable. Not because the patterns are exotic, but because they're familiar. If you've worked in procurement long enough, you've seen at least some of these and probably assumed they were coincidence.
The four forms of bid rigging
Before getting into detection, it helps to understand what you're looking for. Bid rigging generally takes four forms:
Cover bidding. Companies submit bids designed to lose. They bid too high, or include conditions they know will disqualify them. The purpose is to create the appearance of competition while ensuring a designated winner. This is the most common form because it's the hardest to prove — the bids are real, they just weren't meant to win.
Bid rotation. Companies take turns winning. Company A wins the first tender, Company B wins the second, Company C wins the third. Over a multi-year period, each company gets roughly equal share. The pattern only becomes visible when you zoom out across many tenders.
Bid suppression. One or more firms agree not to bid, or to withdraw their bid after submission. This reduces competition for the pre-selected winner. When a firm that regularly bids on certain types of work suddenly stops appearing, it might be a business decision. Or it might be an agreement.
Market allocation. Competitors carve up the market geographically or by sector and agree not to encroach on each other's territory. "You take Riga, I take Liepaja, he takes Daugavpils." Each firm appears to be a local specialist. In reality, they've divided the map.
The red flags nobody wants to see
The OECD checklist identifies specific patterns. Here are the ones we think matter most:
Pricing patterns that don't make sense. When all bids cluster suspiciously close together, or when losing bids are always exactly 5-10% above the winner. In competitive markets, pricing should reflect different cost structures, different approaches, different margins. Uniform price gaps suggest coordination.
The same firm always wins in a specific category or region. Competition should produce variety. If the same company wins every IT tender in a specific city for five years running, and other capable firms exist in the market, the question is why.
Unusual subcontracting patterns. The losing bidder becomes the winning bidder's subcontractor. This can be legitimate — procurement is a small world. But when it happens repeatedly across multiple tenders, it starts to look like compensation for deliberately losing.
Identical errors or formatting. When multiple bids contain the same unusual formatting, the same typos, or references to the same internal document numbers, it suggests they were prepared together or by the same person. We've seen this in documents we analyze — similar phrasing patterns that go beyond industry standard language.
Bid withdrawals timed to benefit a specific competitor. A firm bids, then withdraws at the last moment. The remaining competitor wins by default. Once is circumstance. Twice is pattern.
What's happening in detection right now
Several countries are deploying AI specifically for bid rigging detection, and the results are striking.
Spain's competition authority launched BRAVA in 2024 — a machine learning system that classifies bids as potentially collusive or competitive based on patterns in the national procurement database. It processes thousands of tenders and flags statistical anomalies that no human analyst could spot across that volume.
The UK's Competition and Markets Authority (CMA) started trialing an AI collusion detection tool in January 2025. Even the pilot with a single government department produced actionable results.
Brazil's ALICE system has suspended R$9.7 billion in suspicious procurement transactions. Ukraine's Dozorro — operating under wartime conditions — achieved a 298% increase in collusion detection across 3,000+ daily tenders.
And in September 2024, the OECD launched an EU-funded project to help six member states improve bid rigging detection. In 2025, a second project launched covering six more countries — including Latvia. The fact that Latvia is in this group signals both an acknowledgment of the problem and a commitment to address it.
The data is right there
Here's what I find interesting about this space: the data to detect bid rigging already exists. National procurement databases contain years of bidding history — who bid, at what price, who won, who subcontracted to whom.
The patterns are in the data. The problem was always analysis capacity. A human analyst looking at individual tenders sees individual tenders. A system analyzing thousands of tenders simultaneously sees the patterns that only emerge at scale — the rotation cycles, the geographic carve-ups, the suspiciously consistent price differentials.
This isn't about catching individual corrupt officials (though that happens). It's about identifying structural patterns that distort competition and waste public money.
Where evaluation meets detection
Our focus is on bid evaluation, not fraud detection. Different problem, different tools. But there's an overlap worth noting.
When an AI agent reads every bid thoroughly and documents every finding with evidence, it creates a data trail that makes evaluation more transparent. Bid rigging works in the dark — in the gap between what's in the documents and what gets reviewed. When nothing goes unread, that gap shrinks.
A cover bid — deliberately non-competitive — still contains signals. Vague language where other bids are specific. Missing details that a serious bidder would include. Pricing that doesn't align with the proposed scope. An AI that reads and compares all bids equally will note these discrepancies even if it doesn't know they're intentional.
We're not claiming our tool detects bid rigging. We're saying that thorough, equal, documented evaluation of all bids creates a transparency layer that makes rigging harder to hide.
The uncomfortable math
Bid rigging accounted for 44% of all cartel infringement decisions in the EU in 2023. The estimated cost of corruption risk in EU procurement between 2016 and 2021 was EUR 29.6 billion.
For context: EUR 29.6 billion is roughly the combined annual procurement spend of all three Baltic states. It's enough to build several new hospitals, or fund the entire Latvian education system for years.
Every euro lost to bid rigging is a euro not spent on public services. The detection tools exist. The data exists. The question is whether the political will exists to use them systematically.
From what we're seeing — with OECD projects, national AI tools, and EU attention on procurement integrity — the answer is shifting toward yes. Slowly, but measurably.