Copy article

Algorithmic pricing is becoming a collusion risk. The UK CMA is warning boards, not just data teams

ended 10. March 2026

The Competition and Markets Authority has put algorithmic price collusion on the agenda, warning that increasingly powerful AI can produce coordinated outcomes that raise prices without a smoke-filled room. That is a shift in tone: pricing is no longer “commercial optimisation”. It is a governance and legal exposure that can sit in a model, a vendor tool, or a shared data hub.

The uncomfortable point is that compliance cannot be bolted on after deployment. If an organisation cannot explain how a pricing system learns, reacts, and uses data, it is effectively delegating competition risk to code. A sophisticated model can soften competition through predictable reactions and rapid matching, even when humans never talk.

That makes board-level questions unavoidable. Who signs off the objective function? Who audits whether the model is using rival signals indirectly? What evidence exists when regulators ask how a price was set at 09:12 on a Monday?

This is also a consumer trust issue. Rapid price shifts can feel like a trap, particularly for vulnerable customers. “Dynamic” becomes “unaccountable” when the system cannot be interrogated.

Questions for comment:

  • Should algorithmic pricing systems have mandatory audit trails linking decisions to model version, inputs and constraints?
  • Where should liability sit when third-party pricing tools facilitate illegal outcomes?
  • How should regulators distinguish competition harm from legitimate demand-based pricing?
  • What tests would prove a system is not learning collusive behaviour?
  • Is transparency to consumers enough, or is a hard limit needed on personalisation and speed?

2 responses from the Newspage community

Copy all

Star Quote
Copy

Rohit Parmar-MistryVerified
Founder at Pattrn Data
Algorithmic pricing has moved from ‘commercial optimisation’ to board-level liability. If your model learns that matching rivals fast keeps margins stable, you can drift into a coordinated outcome without any smoke-filled room. The CMA is signalling it will not accept ‘the algorithm did it’ as a defence.

The fix is not a bigger policy PDF. It is evidence. You need an audit trail that ties each price to: model version, objective function, key inputs, constraints, and any human override. Then stress-test for collusion patterns: does the system reward price leadership, punish undercutting, or infer competitors’ intent from shared signals and data hubs?

If you buy a third-party tool, governance still sits with you. Contracts should specify testing, logging, and access for regulators. Transparency to customers helps, but it is not compliance. Until you can explain why a price changed at 09:12 on Monday, you are delegating competition risk to code.
Copy

Editor
Kate AllenVerified
Owner at Finest Stays
The holiday lettings sector is full of agencies using third party dynamic pricing tools that they often cannot properly interrogate themselves. The inputs tend to be basic such as minimum and maximum pricing, seasonality rules and length of stay settings, but very few operators can actually explain how the pricing engine works in its rawest form. Data is pulled from multiple sources including competitors using the same software and pricing scraped from booking websites, which means prices can change rapidly across the market with little human understanding of what is driving it.

We use analytics through Key Data to analyse pacing, occupancy and rates, but pricing decisions are ultimately made manually using our knowledge of the local market. Dynamic pricing may be easy to deploy, but if businesses cannot explain how those algorithms behave it raises real questions about transparency, accountability and consumer trust.