The missing link in personalized-pricing evidence

Development dated October 28, 2026

The FTC’s latest inquiry into personalized prices and worker compensation brings a specific evidentiary problem into focus. Knowing that a business collects personal data, or that two people receive different offers, does not establish how one caused the other. The connection lies in the decision: which information entered the calculation, how it was used, and what offer resulted. [1] [2]

The Commission has ordered information from nine businesses, including airlines, retailers, ride-sharing and delivery platforms. The special-report order rests on an October 28 resolution and covers personal data used in consumer pricing, job assignment and worker compensation. This is a study using the FTC’s compulsory information-gathering authority, not a finding that the recipients engaged in unlawful conduct. Its eventual results remain unknown. [1] [2]

The order asks for more than an inventory of data. For pricing methods, it seeks the inputs, whether they vary across people or groups, the steps using each input, treatment of missing values, and circumstances in which a price is overridden. It also asks about testing and validation. Read together, those requests follow the path from information to an actual decision, including the exceptions that a general description of an algorithm can leave out. [2]

An earlier New York inquiry shows why that detail matters. In a January 8, 2026 letter to Instacart, the Attorney General’s office distinguished reported price variation from the explanation for it. The letter recorded the company’s account of random assignment to price-testing groups alongside statements about behavioral data and promotions. It requested the selection rules, adjusted variables, use of customer data, and effects on prices paid. That letter was an information request expressing concerns, not an adjudicated finding. The question is how customers entered the groups and how their data affected the offer; price differences alone cannot resolve it. [3]

The alternative explanations also matter to the policy argument. Canada’s Competition Bureau summarized competing views in its January 2026 consultation report: some respondents described efficiency gains, lower off-peak prices and targeted discounts, while others raised concerns about opacity, discrimination and competition. The report explicitly says it summarizes submissions and does not necessarily reflect the Bureau’s views. It supplies neither proof that personalized pricing generally benefits consumers nor proof that it generally harms them. [4]

These documents point toward a more precise account of data use. A browsing record could be stored without affecting a particular offer. A price could change because of demand, timing, an experiment or information about the individual. These are possible mechanisms, not findings about the companies under inquiry. Distinguishing them requires evidence linking the relevant inputs to the decision process. A broad label such as “AI pricing” leaves that relationship unresolved. [2] [3] [4]

The inclusion of worker compensation gives the question another dimension: the same inquiry can examine what someone is charged and what someone else is offered for doing the work. Knowing a consumer’s price does not, by itself, explain how a worker’s offer was set. For this new study, the consequential evidence will be whether and how personal information influenced each decision, under which conditions, and with what effects. The order opens that investigation; it does not supply its answer. [1] [2]