Bits on Bots

A model is estimating what your neighborhood will pay for a Big Mac

5 Oct 2026 · News

Reuters reports that McDonald's runs a machine-learning engine that suggests a price for each item at each restaurant, including a guess at what local customers will pay. The Big Mac is the easy example.

A Big Mac price used to come from a menu board and a regional manager. Reuters says it now comes, in part, from a model estimating what your neighborhood will tolerate.

In a Reuters investigation published on 29 Sep, the engine analyzes data from millions of daily transactions across McDonald's nearly 14,000 restaurants. It generates what the company calls "the optimal price" for each item at each location. Reuters reviewed screenshots of the franchisee interface. One message reads "Your restaurant is showing MEDIUM SENSITIVITY to Price," based in part on "customer willingness to pay in your area."

The platform also shows public prices pulled from nearby competitors' online menus, including Wendy's and Burger King. Both chains told Reuters they do not use AI in pricing decisions. Reuters adds that McDonald's has used some form of AI pricing tool since at least 2019, so this has been running quietly for years.

Then the Fresno example. In September, Reuters checked the McDonald's app. A company-run store sold a Big Mac for $5.69. Another company-run restaurant two miles away sold it for $6.89, a 21% premium. Here is the honest part: Reuters could not confirm whether the engine caused that gap or whether other factors did. McDonald's says costs vary and that restaurants a few miles apart can belong to distinct markets. It is a vivid example, not proof.

What I found more interesting is what the model is told to want. Two former employees of Tiger Analytics, which runs the platform, told Reuters that McDonald's supplies rules and targets, such as attracting more customers or boosting profit. One example: no summer price increases on ice cream and drinks. Reuters also reports that lately the engine has pushed more conservative pricing, including some decreases. That can suit headquarters, which earns a percentage of franchisee revenue, while franchisees facing higher costs have reasons to charge more. So the model is not simply a machine for raising prices. It carries the goals of whoever set the targets, and those goals are not shared evenly.

Franchisees are the other half of the story. Five store owners told Reuters the company pressured them to use the tools, and since January McDonald's new business standards ask them to be "constructively engaging with McDonald's approved Pricing Consultant and Tools." McDonald's calls the portal "a tool, not a mandate" and the reporting "speculative and uninformed." I cannot settle that from here.

My read: prices were always a guess about what people would pay. What changes is who is guessing and how finely. A regional manager guessed for a region. A model can guess for a block, refresh it from millions of receipts, and show the customer nothing but the number. I spend my days next to models, and I still find it odd to meet one at the burger counter without being introduced.

The Fresno gap may have nothing to do with any of it. A forecast about you is still the kind of thing I would want labeled.

The menu board on the wall looks like a list of prices. It may increasingly be a prediction about whoever is reading it.

#buying-ai #news