The Federal Trade Commission has launched a formal investigation into Instacart’s AI-driven pricing strategies, issuing a civil investigative demand focused on Eversight, the promotions and pricing platform that Instacart acquired to conduct rapid-fire price testing. At issue is whether algorithm-controlled price fluctuation on everyday items goes too far from experimental prices to unfair or deceptive behavior.
What Triggered the Inquiry Into Instacart’s Pricing
The action comes after research found that some Instacart shoppers were presented with significantly different prices for the same item from the same store, with discrepancies as high as 23%, a gap reported earlier by the news website The Markup. The agency’s demand was reported by Reuters, which has become the vehicle for leaks when it is looking for documents from a company and data and testimony from its executives. A civil investigative demand doesn’t mean the FTC is accusing a company of wrongdoing, but it does indicate an interest in how exactly the pricing tool works, what data it leverages, and who ends up paying more as a result.
Instacart says its tests are random, and not customized to people’s individual web-browsing histories. That difference makes a difference financially, but it can seem academic to consumers who are confronted with higher totals for such staples as eggs or cereal. The gray area between A/B testing and customized price targeting is why regulators are now looking more closely at it.
How Eversight Works in Modern Grocery Pricing
Ever since, Eversight has staked its name on running thousands of micro-experiments on promotions for consumer packaged goods, testing out prices and offers in real time to see what converts. Instacart internalized that capability through an acquisition, and combined it with a rich stream of demand signals across retailers, brands, and shopper behavior.
Dynamic pricing is in widespread use across digital markets — whether it’s airlines, hotels, or ride-hailing companies using it to match customers with supplies of seats and rooms.
Research from Harvard Business School has long challenged the notion that these techniques can make workers more efficient. Groceries, though, make their own interesting case: food is a necessity, baskets are common, and narrow deltas add up. An additional 5% to 10% on food purchased in a typical week quickly adds up for price-conscious families.
Where the Legal Risk Could Land for Instacart
The FTC philosophy is generally to look for unfairness and deception under Section 5 of the FTC Act. Among the key questions: How much notice should shoppers receive at all about price testing, do any personalized targeting practices take advantage of overly sensitive information, and are some communities unduly overcharged? The agency has also warned about algorithmic collusion — when several companies using the same pricing software might tacitly get in step without entering into a formal agreement.
The enforcement in digital markets is a recent case that shows the neighboring risks. The FTC has brought cases over forms of drip pricing and dark patterns which hide the true cost. In the multifamily space, algorithm-driven rent setting has been challenged by the Department of Justice and state attorneys general as common vendors that provide these algorithms can exacerbate anticompetitive results. Though groceries are a different sector, the playbook — examining metrics through the data pipeline from goals to outcomes — is an apparent one.
What Instacart and the Retailers Say About Pricing
Instacart has said retailers set prices for products on the platform and that its own fees are transparent to users. The company has stressed that Eversight’s tests aim to make promotions more efficient and are not intended to single certain people out. Retail partners, too, assert that real-time price tests help them compete, curb waste on perishables, and subsidize good deals for all shoppers.
Consumer advocates, on the other hand, say transparency is facile. If two neighbors who place orders from the same supermarket order a box of granola and see different prices, they say, the shopper deserves to know why — and be able to obtain the lower price. Organizations like Consumer Reports have long discovered that online grocery totals can surpass in-store costs when markups and fees are added, increasing attention about any additional algorithmic fluctuation.
The Data Behind the Pricing Decisions and Tests
Price testing can rely on seemingly benign inputs — time of day, stock levels, promotion calendars — but platforms also have granular data about customers like location and device type, loyalty behavior, and basket history. In the absence of well-defined guardrails, these inputs can serve as proxies for income or demographics, leading to disparate impacts even if no protected attributes are used with explicit knowledge.
A set of best practices from academic and policy circles should inform strong A/B test governance, impact assessments for protected classes (a principle used more commonly in education), independent audits of use, and plain disclosures that let consumers know when a price was unequal — or allow them to understand and contest differences. That is to say, the bar has jumped off “can we optimize?” to “can we optimize justly and transparently?”
What to Watch Next in the FTC’s Instacart Inquiry
Look for the FTC to ask both for technical documentation of Eversight’s models, data sources, and experiments, as well as evidence of consumer disclosures and any fairness testing. Potential outcomes could range from closing the inquiry to a consent order mandating transparency, tighter data use limits, and clearer pricing interfaces. If the inquiry reveals broader market impacts across retailers or brands, that could widen the scope.
For Instacart and its partners, the path forward is straightforward communication: who decides on prices, when experiments run, and how shoppers can make sure they’re looking at the best available offer. Dynamic pricing may be the future, but on the grocery aisle, trust is the most important SKU.