Algorithmic Collusion
What is Algorithmic Collusion?
Algorithmic collusion is price coordination between competitors that software enables, sustains, or produces, even without a direct human agreement. For example, it covers repricing cartels, rivals sharing one pricing engine, and self-learning algorithms that drift toward higher prices. So the term names an antitrust risk, not a pricing technique.
For technology service providers, the exposure sits where prices get set: price books, discount rules, and the quote itself. That matters for MSPs and XaaS sellers that buy outside pricing tools or benchmark feeds. Canada's Competition Bureau addressed shared data in a June 2025 discussion paper. Pooling data from multiple competitors, it said, could make an algorithm's recommendations resemble collusive outcomes.
Synonyms
- Price fixing by algorithm
- Hub-and-spoke collusion
- Autonomous tacit collusion
- Tacit coordination
Why Algorithmic Collusion Matters
- Same law: The FTC's March 2024 business guidance says price fixing through an algorithm is still price fixing.
- Recommendations count too: That same FTC post says software that recommends rather than sets a price is not automatically legal. It also says that agreeing to use an algorithm is itself an agreement.
- Shared vendors: The UK CMA's 2021 paper on algorithms also looked at shared tools. It warned that rivals using the same third-party pricing software can form a hub-and-spoke structure.
- Open questions: That CMA paper said it was as yet unclear whether authorities can object to hub-and-spoke or autonomous tacit collusion without direct contact between rivals. As a result, buyers of pricing tools carry uncertainty as well as risk.
Close bigger deals with servicePath™ CPQ+. See how governed, AI-native quoting works in practice. Book your demo
Key Features of Algorithmic Collusion
- Automated cartels: Rivals agree on prices, then use software to keep the deal in place. The CMA also noted that automation makes deviations easier to detect and answer.
- A closed UK example: The CMA's online posters case 50223 closed in 2016. In it, two Amazon UK sellers used automated repricing software to enforce an agreement not to undercut each other. Trod Ltd accepted a fine of £163,371, while GB eye was not fined under the leniency policy.
- A common hub: Several competitors route pricing through one vendor or intermediary. Canada's Competition Bureau notes that third-party algorithm suppliers may act as the hub in hub-and-spoke agreements.
- Platform messages: In Eturas (Case C-74/14), decided on 21 January 2016, the Court of Justice of the European Union (CJEU) considered a booking-system message capping discounts. It held that agencies aware of the message could be presumed to join a concerted practice. Still, they could rebut that, for example by publicly distancing themselves or reporting it to the authorities.
- Autonomous learning: In simulations, self-learning algorithms have reached higher prices with no communication at all. For example, in a 2020 American Economic Review study, Q-learning pricing algorithms in a simulated market consistently learned supracompetitive prices.
- No meeting to point to: Coordination can emerge without an email, call, or meeting between rivals. As a result, it is harder to spot inside a sales team and harder for authorities to prove.
How Algorithmic Collusion Works
The OECD's 2023 background note on algorithmic competition describes three routes: explicit agreements, hub-and-spoke settings, and autonomous tacit collusion. The steps below show a typical hub-and-spoke pattern. But not every case follows every step.
- Adopt shared software: First, several competitors license the same pricing tool or hand pricing to one intermediary.
- Feed in data: Then the tool takes in each user's prices, costs, or demand signals, sometimes alongside rivals' data.
- Generate recommendations: Next, the algorithm proposes prices for each user from the combined picture.
- Monitor rival prices: Meanwhile, software watches competitor prices continuously and flags any undercutting.
- Answer price cuts: Discounts trigger fast matching responses, so undercutting stops paying off.
- Settle at higher levels: Finally, prices stabilize above what independent competition would produce.
Algorithmic Collusion vs. Price Fixing
People confuse the two because both end in coordinated prices, but the mechanism and the evidence differ.
| Dimension | Algorithmic Collusion | Price Fixing |
|---|---|---|
| What it is | Coordination on prices that software enables, sustains, or produces, explicitly or tacitly | An agreement among competitors to raise, lower, maintain, or stabilize prices |
| How coordination happens | Shared software, pooled data, automated monitoring, or self-learning algorithms | A written, verbal, or inferred agreement between rivals |
| Role of human agreement | Direct, indirect through a common vendor, or possibly absent | Required, though it can be inferred from conduct |
| Legal position | Treated as price fixing where software enforces an agreement; regulators have called purely autonomous outcomes unclear | US antitrust law generally requires each firm to set prices on its own, without agreeing with a competitor |
| Where they overlap | Algorithmic price fixing: an ordinary cartel run through code | Same conduct, judged under the same price-fixing rules |
The FTC's guide to price fixing defines it as an agreement that can be written, verbal, or inferred from conduct. Yet the guide says a firm may charge the same price as rivals if no agreement drove the decision.
Common Challenges
- Vendor opacity: A pricing tool buyer may not know whose data feeds the model, or which rivals use it.
- Black-box logic: Canada's paper flags the "black box issue", because AI-driven pricing can be hard to explain.
- Thin evidence, open law: The OECD's 2023 note said there was not conclusive evidence that autonomous tacit collusion is a significant issue. Still, the OECD's 2017 paper asked whether agencies should revise the concepts of agreement and tacit collusion.
- Suggestions that feel optional: Reps may treat a suggested price as advice. But the FTC says recommendations can still form part of an unlawful agreement, and price deviations do not cure it.
Best Practices
- Price from your own inputs: Build prices from your own costs, customer value, and commercial rules. A value-based pricing model anchored in your own economics is easier to explain than a feed that blends rivals' data.
- Question every vendor: Ask what data a tool uses, whether it pools non-public competitor data, and how it produces recommendations. This applies to algorithmic pricing and predictive pricing tools alike.
- Document the rationale: Record why each price or discount was set. For example, a short note tied to cost or value helps show you set the price independently.
- Keep a human approver: Send out-of-policy prices to a named approver instead of auto-accepting software output. Strong deal governance makes this routine rather than an exception.
- Escalate coordination signals: Forward any vendor or platform message that hints at aligned pricing to your legal team at once.
This page is general information, not legal advice, so consult antitrust counsel about any specific situation.
How servicePath™ Helps
servicePath™ CPQ+ does not replace legal review, but it supports pricing built on your own rules, approvals, and records.
- Your own pricing logic: servicePath™ CPQ+ lets teams build business rules, configured solutions, and pricing into the platform. So quotes can follow your own deterministic pricing logic rather than an outside engine's suggestion.
- Cost and margin clarity: servicePath™ CPQ+ gives real-time visibility into deal financials. It also clarifies cost of service and unit of service, so sellers can anchor each price to their own economics.
- Threshold approvals: Approval workflows and conditional logic route deals and discounts to the right approvers, based on predefined thresholds.
- Governance and audit trail: Custom governance and legal checkpoints sit inside the quoting logic, while built-in audit trails record decisions. In servicePath™'s view, AI can assist the workflow, but approved rules and human authority determine the committed price.
Frequently Asked Questions (FAQs)
Does antitrust law apply to algorithmic collusion?
Yes, where software carries out a price agreement, which the FTC still calls price fixing. Regulators have described purely autonomous algorithmic coordination as a less settled area. Speak with antitrust counsel about specific situations.
Can using the same pricing software as competitors create risk?
Yes. Canada's Competition Bureau says using common pricing algorithms or pooling data among competitors may raise issues under the Competition Act. A shared vendor can act as the hub that links rivals' prices.
Can AI pricing algorithms collude without being told to?
In experiments, yes. A 2020 American Economic Review study found Q-learning pricing algorithms consistently learned supracompetitive prices without communicating. Still, the OECD and the UK CMA have since called real-world evidence limited.
How can B2B sellers reduce algorithmic collusion risk?
The defensible pattern is pricing from your own costs, value, and rules. So teams question tools that pool rivals' non-public data and document each pricing rationale. Require human approval for exceptions, and ask antitrust counsel about specific arrangements.
Related Terms
- Algorithmic Pricing
- Deterministic Pricing
- Competitive Pricing
- Deal Governance
- Pricing Strategy
- AI Price Optimization
- AI Governance
- Price Elasticity
Ready to take the next step?
Talk to a servicePath™ CPQ Architect about building governed, AI-native quoting with the right controls underneath the agent.
Book a conversation with servicePath™