McDonald's tracks franchisee deviations from its per-restaurant AI menu prices

McDonald's sets both the objective and the guardrails. According to two former Tiger Analytics employees, the company works with the platform's developer and sets its business goals, such as drawing more visitors or raising profit ^[13]. The rules they described also include concentrating increases on items whose price has not moved in at least two years. Another rule allows increases only on items that already cost more in at least 30% of restaurants ^[14]. Tiger Analytics did not comment ^[15].

The 30% rule is the one I would ask about at review. If it means what it appears to, a store can raise an item only where a sizeable share of the network already charges more. The fleet's own price distribution then caps how far any single store can lead ^[14]. The output ships in batches. Franchisees have received recommendations at least three times a year since 2023 ^[12]. That was the year chief executive Chris Kempczinski told investors McDonald's had built its own tools to assess prices at each restaurant ^[12].

Who sets the goal matters because the two parties are paid differently. McDonald's takes most of its revenue as a share of restaurants' total sales, whatever a single store earns, so cheaper menus that pull in more customers can lift its take ^[16]. The owner pays wages and rent out of the margin. The National Restaurant Association estimates those costs are up 36% since 2019 ^[17]. The output has swung both ways. Some owners objected to recommendations for steep increases during and after the pandemic, and in recent months the system has proposed restraint, including some cuts ^[18]. In September McDonald's told investors the platform is an important part of its value strategy, and Kempczinski said franchisees also see that cheap items draw low-income customers ^[19]. Placer.ai estimates McDonald's US traffic has fallen year over year every month since March ^[20].

McDonald's says the platform only recommends and franchisees set the final price ^[8]. Five owners said the company pushed them to use the tools ^[9]. A June company document shows it records store-level deviations from the recommendations in detail ^[10]. A suggested price with a deviation log behind it is a default with an escalation path. Former owner Karen King and other franchisees described calls from company representatives when their prices did not match the platform's, and King said owners have less and less room to set prices themselves ^[11]. McDonald's did not respond to her specific allegations ^[11].

Per-restaurant prices mean gaps between neighbours are expected output. The Fresno pair franchisees cite is $1.20 apart on a Big Mac, or 21% ^[1]. It has not been established whether the algorithm produced that gap or something else did ^[7]. McDonald's says costs and market conditions differ by restaurant, and stores a few miles apart can sit in different markets ^[8]. For the model to be right in Fresno, customers at the two stores have to be willing to pay different amounts. That estimate is what the system produces from purchase data ^[1][2]. At one extreme, in 2023 it suggested franchisee George Mitchell charge about $18 for a Big Mac meal at a restaurant by a Connecticut toll highway ^[21].

The competitor feed reaches outside the company. Wendy's and Burger King, whose online menus the system reads, both said they do not use AI to set prices ^[3][4]. Reuters, in reporting relayed by mezha.net, lists antitrust attention as a risk of McDonald's approach, alongside alienating customers ^[5]. The report does not tie that risk to the competitor feed or describe any regulatory action ^[5].