The Federal Trade Commission has found that pricing intermediaries use a wide range of personal data to set individualized prices for consumers, including grocery shoppers. At the same time, new academic research reveals that the AI systems grocery chains rely on to manage inventory carry built-in forecasting errors that can distort supply and inflate costs. Together, these developments suggest that algorithmic tools are reshaping what Americans pay for food in ways that are difficult for shoppers to detect, or challenge.
Your Data Sets Your Price at the Register
Grocery pricing has long been shaped by regional competition, seasonal supply, and wholesale costs. But a growing layer of algorithmic intermediaries now sits between those traditional forces and the price a shopper actually sees. The FTC’s Surveillance Pricing Study, released in January 2025, found that these intermediaries draw on granular consumer data, including location, browsing patterns, and even mouse movements, to tailor prices and promotions for individual buyers. The study identified at least 250 clients served by these firms, with grocery among the sectors where personalized pricing is actively deployed.
The mechanism works roughly like this: a shopper’s digital trail, from loyalty app usage to the way they scroll through an online store, feeds into pricing models that estimate willingness to pay. The result is that two people buying the same box of cereal or carton of eggs might face different prices without ever knowing it. The FTC study did not quantify exact price differences across grocery categories, and no retailer has publicly disclosed how much individual prices diverge. That gap in transparency is precisely what regulators are trying to close, since consumers cannot meaningfully consent to or contest a pricing system they cannot see.
FTC Orders Target the Firms Behind the Algorithms
The agency’s concern did not begin with the January report. Over the summer of 2024, the FTC used its 6(b) authority to compel information from eight companies about how their surveillance pricing products work and who uses them. The targets included Revionics, Mastercard, Accenture, Bloomreach, PROS, and McKinsey, a roster that spans retail technology, consulting, and payment processing. The breadth of that list signals how deeply algorithmic pricing tools have penetrated the food retail supply chain, from farm gate logistics all the way to the checkout screen.
For everyday shoppers, the practical consequence is straightforward: the price on a digital shelf may reflect not just what a product costs to stock but also what an algorithm calculates a specific buyer will tolerate. Because these systems operate behind retailer platforms, consumers have no way to compare their personalized price against a baseline. The FTC’s orders sought technical details about implementation, data inputs, and client relationships, but the agency has not yet announced enforcement actions tied specifically to grocery surveillance pricing. That leaves a regulatory gap between identifying the practice and stopping it, even as more retailers experiment with dynamic offers and individualized discounts.
AI Forecasting Errors Ripple Through Fresh Food Aisles
Pricing algorithms are only one side of the problem. The AI systems that grocery chains use to predict demand and manage inventory carry their own distortions, and those errors directly affect what ends up on shelves and at what cost. A research paper published on arXiv introduced FreshRetailNet-50K, a stockout-annotated dataset built from large scale hourly store product time series. The dataset was designed to address a specific blind spot: when a product sells out, the sales record shows zero units moved during the stockout period, not the actual demand that went unmet. AI models trained on that incomplete data systematically underestimate true demand for fast-moving items.
This “censored demand” problem hits fresh retail hardest. Perishable goods like produce, dairy, and meat have narrow windows for restocking, and forecasting errors compound quickly. If an AI model learns from stockout-distorted data, it may recommend lower reorder quantities, triggering a cycle of recurring shortages. Alternatively, retailers who recognize the bias may overcorrect by padding orders, which increases spoilage and waste. Either outcome raises costs that get passed along to consumers, whether through higher prices on individual items or broader adjustments in a store’s pricing strategy to recoup losses from waste. The research team’s dataset is designed to help correct this bias by explicitly annotating stockouts, but adoption across the industry remains uneven, and most grocery chains have not disclosed how they handle censored demand in their own forecasting pipelines.
Instacart’s Pricing Tests Show the Consumer Fallout
The clearest public example of algorithmic pricing gone wrong comes from Instacart. The delivery platform ran a program in which users could see different prices for the same item at the same store, with eggs shown at five different prices depending on the buyer. Instacart ended that program after scrutiny, but the episode illustrated how AI-driven pricing can fracture the basic expectation that a product has one price at a given store. For shoppers who relied on Instacart during the pandemic and continued using it as grocery delivery became routine, the revelation that their neighbor might pay less for the same dozen eggs eroded trust in the platform’s pricing integrity and raised questions about how widespread such tests might be across the sector.
Separately, Instacart agreed to a $60 million settlement with the FTC over allegedly deceptive advertising and fee disclosures. The settlement addressed a pattern of opacity around service fees and delivery charges that made it difficult for users to understand the true cost of their orders. Taken together, the variable pricing tests and the fee disclosure case paint a picture of a company whose algorithmic systems repeatedly obscured what consumers were actually paying. While Instacart is the most visible example, it operates within the same ecosystem of pricing intermediaries and demand-forecasting tools that the FTC study flagged across the broader grocery sector, suggesting that similar practices could be embedded in less scrutinized platforms and retailer apps.
Efficiency Gains Exist, but Who Benefits?
Proponents of AI in food supply chains point to real gains. A systematic review in the journal Food Policy examined how machine learning is being deployed in agriculture, logistics, and retail, finding that advanced models can improve yield predictions, optimize routing, and reduce waste throughout the chain. The authors concluded that, in principle, data-driven tools can help stabilize supplies and lower operational costs, especially when combined with better monitoring of perishables and more responsive ordering systems. For grocers, those efficiencies promise fewer empty shelves, more accurate promotions, and less food discarded before it can be sold.
The open question is how those gains are distributed. If forecasting improvements are used primarily to fine-tune personalized prices and extract more revenue from shoppers deemed less price-sensitive, the technology may widen gaps between consumers rather than narrowing them. Conversely, if retailers pair better demand models with transparent pricing policies and clear disclosures about fees and discounts, some of the savings from reduced waste and smoother logistics could flow back to households in the form of more stable prices. The tension between these paths underlies current regulatory debates: the same algorithms that can make food systems more efficient can also make grocery bills more opaque, leaving consumers to shoulder the risks of error while companies capture most of the rewards.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
