Algorithmic Price Fixing and Information Exchange: The Agreement Behind the Tool

Algorithmic price fixing uses pricing software to carry out an agreement among competitors about what customers will pay. The agreement can appear in instructions to a developer, commitments to share confidential pricing data, or decisions to follow a common pricing system. An engineer may know what went into the model. A pricing employee may know why its recommendation became the price. The messages connecting those decisions can explain whose interests the software was built to serve.

Common Pricing Vendors and Nonpublic Rival Data

The antitrust concern with a common pricing vendor is how competitors supply and use nonpublic rival data to coordinate their decisions.

The government’s cases against RealPage and Agri Stats describe different ways that exchange can work. RealPage supplies software that recommends rental prices. Agri Stats supplies reports and consulting that meat processors use to make pricing and production decisions. In both civil cases, the government alleged that competing businesses used confidential information supplied through a common provider to weaken competition.

The information’s detail and timing matter together. A report identifying a particular product, plant or customer can reveal more than an industry average. Older information can still describe a deal that remains in effect. The Justice Department’s explanation of the Agri Stats settlement makes that point with contracts renewed only periodically: the age of the spreadsheet and the age of the commercial decision are different things.

Access matters, too. In the Agri Stats allegations, sellers could see information that the grocery stores and restaurants negotiating with them could not obtain. The market intelligence served one side of the negotiation.

Poster Sellers Agreed on Prices Before They Used the Software

In a criminal plea agreement filed in 2015, a poster seller admitted a conspiracy in which competing sellers agreed on prices and implemented that agreement through pricing algorithms. The posters were sold through Amazon Marketplace.

The sequence is the point. The sellers agreed on pricing, then used the software to carry out what they had decided. For an engineer who received competitor instructions, the configuration history can connect a commercial conversation to the rules that actually changed prices.

RealPage Connected Rival Data to Rental Prices

The Justice Department alleged that RealPage and participating landlords agreed to share competitively sensitive information and use software designed to align rental pricing. Its November 2025 settlement explanation describes both the data entering the system and the features influencing the prices coming out.

The inputs went well beyond advertised rents. The government described rental applications, executed new leases, renewal offers and acceptances, and occupancy estimates and projections. AI Revenue Management and YieldStar drew information from property management systems used to collect rent, manage leases, handle maintenance requests and perform accounting. The records of operating an apartment business became inputs for pricing a competing property. The government alleged that renters lost the benefit of landlords competing on price.

The government also alleged that competing landlords discussed pricing strategies and market trends based on nonpublic information at RealPage software meetings. The data connection and the conversation connection belonged to the same alleged arrangement.

Training Data and the Price Recommended Today

A rival’s confidential information can influence both model training and the live price recommendation.

The proposed judgment described in the RealPage filing restricts the use of competing landlords’ nonpublic data when generating prices. Its training restrictions permit certain nonpublic data only after it is at least 12 months old and no longer comes from active leases.

For the engineer, that separates questions often collapsed into a single statement about using historical data. Which model received the information? Did it describe an active lease? Did it train a model, or directly help set the next recommended price? The source fields and the model’s use of them tell the story.

Automatic Acceptance and the Ability to Cut Prices

The government alleged that RealPage’s pricing controls favored increases: its software would not recommend prices below a competitor-derived floor, yet allowed recommendations above a ceiling also based on competitors’ data. There was room to go higher and a barrier to going lower.

The proposed terms address that imbalance. They require equal percentage room below the floor and above the ceiling, with the landlord setting the permitted range. They also require users to be able to configure the auto-accept feature individually, protect the ability to reject or override a recommendation, and prohibit requirements or incentives to accept recommended prices.

The filing also describes a sold-out guardrail that increases the recommended rent when a floor plan reaches its target occupancy. The proposed terms bar use of competitors’ data in that feature. They also bar pricing advisors, the employees who advise on software settings and revenue management, from sharing other owners’ nonpublic data while helping set rents or occupancy levels.

Those controls make the acceptance record important. A recommendation, the setting that automatically accepted it, and the eventual rent can show how a system influenced the business. An override request and the response can show who controlled a departure from the recommendation.

Agri Stats Made Competitors’ Products Comparable

Agri Stats allegedly standardized confidential meat-processing data so sellers could compare their own products and operations with those of competitors. The government’s complaint concerns broiler chicken, pork and turkey, with reports covering prices, margins, inventories, costs and production.

The government’s later settlement explanation alleged that Agri Stats assigned product codes using specifications such as cut, weight, packaging and preservation method. A processor could compare its chicken breast with closely matched products from competing processors. The comparison reached almost to the individual product code.

The government alleged that sales reports showed processors where their prices ranked and where they could raise them. Consulting sessions then identified products and customers to target for increases. A sales employee negotiating with a grocery buyer could have a view of competing offers that the buyer lacked.

The information was also checked. According to the second amended complaint, Agri Stats audited submissions for accuracy and completeness, reducing the ability of participants to hide or withhold information from rivals. The alleged service gave competitors confidence in what they were learning about one another.

What Happens When the Data Is Called Anonymous?

Detailed reports can reveal which competitor sits behind an anonymous entry. The Agri Stats complaint alleged that processors recognized distinctive features of plants or products, then tracked the identified facilities across reports.

The complaint described how readily one employee said that could be done:

The government also alleged that Tyson employees from different departments met to identify competing pork facilities in the data. For an insider, the useful connection is the report, the distinctive feature that identified the facility, and the internal discussion using that identity. The label on the report tells only part of the story.

The Reports Reached Production and Bonuses

The alleged exchange reached decisions about supply as well as sales prices. The complaint described processors using rivals’ production plans to monitor whether anyone was expanding output. It also alleged that processors used relative profit-margin reports to make executive bonus decisions, tying compensation to margins compared with competitors.

The government further alleged that individual Agri Stats consultants advised competing processors at the same time. Those consulting relationships connect the information exchange to the people interpreting the reports and recommending what to do next.

The May 2026 settlement explanation adds a concrete finding from the government’s economic expert. In the broiler chicken analysis, processors were nearly twice as likely to raise prices on products the reports identified as relatively low priced as to lower prices on products identified as relatively high priced. The government said that analysis confirmed documents showing processors using the reports to target increases.

That comparison gets to the commercial purpose of the exchange. The question is what the business did after learning where it stood.

The Proposed Restrictions Reach the Reports Themselves

The Agri Stats proposed judgment targets sales reports and consulting based on pricing information, the detail of production and cost reporting, and buyers’ access to the remaining reports.

The restrictions generally replace facility- or company-specific production, cost and labor data with industry averages or averages by quartile. They also require older data and reports that buyers can purchase too.

The restrictions address the features that made the alleged exchange useful to sellers: a close view of rivals’ operations, information recent enough to influence decisions, and access withheld from buyers.

The Records That Connect the Input to the Price

An insider’s account becomes more concrete when the data, instructions and resulting business decision can be connected. Engineers, pricing staff and vendor employees may each hold a different part. We work through those connections with you.

The source system and data dictionary explain what a field actually means. An advertised price, an executed transaction and a renewal offer answer different questions. Product definitions explain whether a report compares like with like. The Agri Stats product categories and RealPage lease inputs show why those distinctions matter.

The exchange records show who supplied the information and who could use it. Subscriber lists, access records and audit workpapers can connect participants to the reports. In a supposedly anonymous system, internal matching work can explain how employees identified the rival behind a row.

Configuration history, recommendation output and acceptance or override records connect the information to the pricing decision. Beside them belong the instructions: messages from a competitor, a vendor’s consulting advice, or a manager’s direction to use a particular setting. In the poster-pricing plea, the human agreement preceded the algorithm. In the RealPage allegations, the inputs and controls operated together.

The eventual price, discount or production decision completes that commercial account. We examine which information was available when the decision was made and who can explain its use firsthand.

How Is Algorithmic Collusion Detected?

Pricing data become more useful when an insider can explain the instruction behind the change, a connection developed in the evidence page.

The 2018 construction-bidding review also proposes linking tender prices and winners with bidder identities, ownership, subcontracting, quantities and materials. The number becomes more informative when it stays connected to the business and transaction that produced it.

For an algorithmic-pricing matter, we apply that lesson by connecting the price pattern to the source data, the participants and the instructions. Someone who understands the product or built the system can explain a relationship that a price series leaves invisible.

A Whistleblower Case Starts With What You Know

The Justice Department and Postal Service’s antitrust reward program can pay eligible whistleblowers a share of qualifying criminal recoveries. We assess how your information, the companies involved and the customers affected fit the program.

When coordinated software prices reach government purchases, we follow the contract and payment representations too. Bid rigging on a government contract involving knowingly false claims for payment can raise a False Claims Act issue. A whistleblower case under the False Claims Act can pay 15% to 30% of what the government recovers.

You Know How the Price Was Chosen

Our whistleblower experts can work with you to connect a competitor agreement to the data feed, pricing setting or approval that carried it out.