Personalized grocery offers now need a pricing review

A personalized discount is now a pricing decision

A retention offer sent to a lapsing grocery shopper now needs a pricing review. Suppose a marketing team uses purchase history to choose a tailored offer: once it changes what the shopper pays, the team shares responsibility for the price. Under Maryland’s Protection From Predatory Pricing Act, HB 895, a defensible campaign needs evidence of the saving and an explanation of how the recipient was selected.

That responsibility starts with understanding surveillance pricing: using data linked, or reasonably linkable, to a person to set different prices for the same product at the same time. Shopping history, location and online behavior can influence the price a shopper sees. Maryland’s law, effective October 1, 2026, restricts using personal data to impose higher individualized prices at covered food retailers and third-party delivery services. Maryland is the first state to restrict surveillance pricing specifically in the food industry.

The information behind those decisions already flows through marketing systems. Customer data platforms, loyalty apps, retail media networks that sell advertising using retailer data, and delivery platforms supply information for selecting offers and displaying prices. A pricing engine determines a charge, an offer engine selects a promotion, and a personalization engine chooses a shopper’s experience. When their outputs affect payment, campaign designers share accountability with merchandising and revenue management.

Because those systems can change what a customer pays, reviews need to follow the decision from its inputs to the displayed price. Algorithmic pricing tools calculate changes, while electronic shelf labels make them easier to display; the display technology alone reveals little about how the price was chosen. The questions are whether personal information increases an individual’s price or prohibited targeting denies access to an advantage. Personalized offers remain possible when the retailer can substantiate their value and meet the applicable rules.

Price differences can damage trust before a legal finding

The business consequences can arrive before those legal questions are settled. Groundwork Collaborative and Consumer Reports investigated prices on Instacart and found different charges for identical products, with the largest reported difference reaching 23%. Backlash developed within days, and Instacart discontinued the pricing technology. eMarketer reported the shutdown and resulting brand damage, alongside Instacart’s position that it had not used demographic data.

Instacart’s response shows how quickly a pricing experiment can become a customer-facing problem. Stopping the technology after the backlash was a concrete business consequence, while questions of statutory violation and discriminatory intent remained separate. An analytics team therefore needs to assess how customers may experience a test even when the company disputes a sensitive-data explanation.

Pricing intermediaries make this concern relevant beyond one platform. Preliminary findings from the Federal Trade Commission’s staff study of surveillance pricing found that intermediaries frequently used precise location and browsing history to target different prices. The intermediaries studied served at least 250 clients, including grocery stores. That count shows the reach of their data-driven services and says nothing by itself about the legality of individual decisions.

Within that market, household finances give shoppers a strong reason to compare prices. Gartner, a commercial research and advisory company, found that 56% of U.S. consumers were spending as though the economy were in a recession or preparing for one. The finding concerns how people cope with economic conditions, which makes price scrutiny relevant without establishing a response to surveillance pricing.

A separate Gartner personalization survey examines what happens after marketing reaches a customer. It found that 53% of customers experienced negative outcomes from personalized marketing; those customers were 3.2 times more likely to regret a purchase and 44% less likely to buy again. These reported associations give personalization teams a reason to look beyond conversion, while leaving the effects of individualized pricing as a separate question.

For pricing, that later experience can include discovering what another shopper paid. Consider a hypothetical checkout line in which two shoppers find different prices for the same carton of eggs. The visible difference may conflict with their expectation of equal treatment, leading to complaints, reputational damage or regulatory attention. Explaining the model’s inputs may leave that expectation unresolved.

The link between perceived data use and trust is central to Josie Gaeckle’s warning. Gaeckle, senior vice president of client insights at SMG, speaks from a commercial customer-insights role, so her emphasis on understanding customer trust also aligns with her company’s business interests. She says: “Customers need to feel that their data is helping a brand serve them better, not being used to charge them more for the same item. That distinction affects more than an individual transaction: it shapes trust and, over time, brand equity.” Her point concerns the benefit customers believe they receive in exchange for their information.

Making a price easier to assess has a longer commercial history. Congress’s Automobile Information Disclosure Act of 1958 required new U.S. cars to carry a window sticker showing the suggested retail price, giving buyers a disclosed benchmark for assessing a dealer’s asking price. That disclosure grew from a compliance obligation into a commercial trust signal. Grocery offers raise their own baseline question: what price gives the shopper a meaningful way to judge the claimed saving?

Defensibility requires a baseline, fair access and state-specific rules

Answering that baseline question is the first of three distinct reviews. A retailer must establish what price the saving is measured against, who can receive the advantage and which jurisdiction governs the decision. Each requires different evidence: a documented discount can still raise concerns about who receives it, while an offer permitted under one state’s individual-consumer scope can receive different treatment elsewhere.

For the saving itself, the reference price is the baseline used to calculate the discount. Subject to exclusions, HB 895 restricts setting a particular consumer’s personalized price using personal data and using that data to charge one consumer more than another. Yet the law leaves “reference price” and a fair price undefined. So a retailer needs a reason for choosing its baseline as well as a record of the number.

A hypothetical offer shows why that justification matters. A retailer could choose an implied regular price above the amount it intends to charge, then describe the charge as a discount. Recording both figures establishes the arithmetic, but the higher figure still needs substantiation. This illustrates a risk in representing a saving rather than an observed enforcement case.

To make that review possible, a retailer can document a reference price for each stock-keeping unit, or SKU, and preserve its connection to every personalized charge. The records should show whether the offer falls below the relevant baseline and retain the basis for choosing it. Approval must then assess that basis, because recordkeeping alone cannot make an arbitrary reference price fair.

Once the saving is established, the next question is who can receive it. HB 895 restricts offers and advertising that use protected-class data to withhold advantages available to other shoppers; the named characteristics include race, religion and national origin. A review limited to the final amount paid can therefore miss a separate problem in how offers are distributed.

That distribution review must also consider proxies: inputs that can substitute for protected characteristics even when those characteristics are absent from the model. ZIP code, device type and shopping patterns can have this effect. Teams need to inspect those features and test whether offer allocation produces different outcomes across protected groups.

Allocation matters even when every personalized charge stays below a reference-price ceiling. The ceiling limits increases above that level, but shoppers can still be denied discounts through prohibited targeting. Testing therefore needs to examine eligibility, access and subsequent redemption alongside the final charge.

The legal treatment of allocation also depends on whether an individual or a group receives the price. According to IAPP’s analysis, Maryland covers food retailers with at least 15,000 square feet of food-selling space, as well as third-party delivery providers. IAPP also reports that language covering groups of consumers was removed from the final bill. Segment pricing consequently falls outside Maryland’s single-consumer prohibition as described, while remaining subject to other applicable provisions and laws.

Applying that distinction requires a record of whether the system determines a price for an individual or a defined group. The team should document how the group is formed and how its members receive a price, then use that decision unit when applying the law. An output labeled a segment offer still needs scrutiny if the underlying process determines prices individually.

New Jersey changes the review because its law covers both individuals and groups. A segmentation design outside Maryland’s single-consumer prohibition therefore needs a separate assessment there. New Jersey also includes a private right of action, allowing private parties to bring claims; EPIC describes this as a first among state surveillance-pricing laws. That enforcement route adds litigation exposure to the jurisdictional review.

Connecticut extends the coverage question in another direction. Stateline describes its law as applying to retailers more broadly than Maryland’s food-focused approach. Stateline also reported enactments in Connecticut, Maryland and New Jersey, alongside consideration of comparable bills by lawmakers in at least 11 states. A national campaign must therefore account for retailer coverage, individual-versus-group decisions and available enforcement routes.

Within Maryland’s framework, exclusions preserve several approaches to personalization. Greenberg identifies exclusions for promotional pricing; loyalty and rewards programs; subscription pricing; geographic and cost-based price variation; changes resulting from consumer-consented data exchanges; pricing-error corrections; and outage-related resets. Hunton also identifies temporary discounts or price changes associated with retaining existing customers. These exclusions leave room for marketing activity, while each campaign still requires review of its other obligations.

Redesign campaigns around verifiable savings and appropriate test units

Those legal distinctions need to shape campaign requirements before an offer reaches customers. For the lapsing grocery shopper, the team must be able to show the saving, explain the selection and apply the rules governing the recipient. The table separates the Maryland treatments and interstate differences described above from recommended operating responses.

Experimental design needs particular attention because randomly assigning shoppers to test groups can place individual consumers in a higher-price condition. A test intended to estimate price sensitivity may therefore create the individualized increase that raises exposure. Store-level, time-based or offer-based designs can reduce that risk, subject to the rules governing the location, group and offer.

To make live campaigns and experiments reviewable, record four fields for every relevant decision: reference price, offer type, consent status and jurisdiction. Together, these explain the claimed saving, the category of treatment, the role of permission and the applicable legal context. Capturing them when the decision occurs reduces the need to reconstruct a campaign after a complaint.

Those records are easier to substantiate when the program’s design makes its benefit clear. Open-enrollment loyalty programs make access easier to establish, retention offers tied to a baseline make savings easier to demonstrate, and regional differences supported by actual costs make variation easier to explain. These designs can reduce exposure when their supporting evidence matches how the program operates.

The same need for proof appears in the perspective of Hemant Warier, head of marketing at Orvera AI. Warier speaks for an AI company with a commercial interest in the market for AI systems, so his comments should be read in that context. He says: “A brand will want to know that each time they provide a retention offer they can prove it’s a discount or loyalty program and no one overpaid based on their data. Human or AI agent.” His expectation covers interactions conducted by people and AI agents, which means campaign records must remain usable across both.

Meeting that expectation changes the actions a personalization model can trigger. Shopper behavior can remain useful for learning, while permitted outputs move toward selecting offers and delivering demonstrable value. Teams accustomed to optimizing margin for each individual need to define those output limits explicitly and evaluate the customer experience they produce.

Require pricing systems and vendors to produce the evidence

Campaign limits become enforceable when the systems carrying them out expose the necessary information and controls. A retailer should be able to inspect incoming data, connect an offer to its baseline, constrain the charge and retrieve the decision record. Internal reviews and vendor assessments need to test those capabilities before campaigns depend on them.

Internal controls

Start with the data entering the decision, because an output review can miss a prohibited input. Inventory every signal entering pricing, offer and personalization engines, including signals from customer data platforms, loyalty apps, retail media networks and delivery platforms. Mark each signal as personal or aggregate and record its consent status, so reviewers can assess both the information used and permission for its use.

Connect that inventory to a reference-price ledger that records baselines by SKU and store. Express each personalized offer relative to the relevant entry, preserving the evidence that distinguishes a discount from an individualized markup. The ledger must also retain the basis for the reference price so an approver can judge whether the claimed saving is meaningful.

Once those records substantiate the saving, they can support customer-facing explanations. A retailer could use them for a pledge about consistent pricing or a public-facing page showing loyalty members what they save. Any promise needs to stay within what the operating controls demonstrate, and its effect on customers should be tested.

For retention campaigns, test that explanation by presenting the offer as a saving relative to the shelf price and comparing it with the current approach. Establish a holdout group that does not receive the tested treatment, then measure redemption, repeat purchase and complaint rates before expanding the campaign. This brings the opening grocery example into the review: the team checks both the response to the offer and the experience after receipt.

Monitor the wider response alongside that experiment. Track app ratings, support tickets mentioning price and cart abandonment after price display, then review them with conversion results. A successful transaction can coexist with dissatisfaction that surfaces later or through another channel.

Use those findings in campaign approval, because an auditable decision can still require a change. A recorded baseline needs justification, and a documented allocation still needs to meet the governing rules. If the records reveal a weak baseline, prohibited targeting or treatment incompatible with a jurisdiction’s rules, revise the campaign before launch; documentation creates neither a compliance safe harbor nor an assurance of preserved profitability.

Vendor questions before renewal

The retailer also needs evidence from vendors whose systems make or influence these decisions. Before renewing a pricing, loyalty or personalization contract, ask which data enters the engine and obtain its complete model-input list. Confirm whether personal data is used and whether the retailer can inspect the inputs that influence decisions.

After establishing the inputs, ask which outputs the system can prevent. Confirm that it supports a hard reference-price ceiling and can stop individualized charges from exceeding it. Keep separate controls for offer eligibility, because a ceiling cannot establish fair access to the savings below it.

Those controls must be explainable after a decision runs. Confirm that the vendor retains decision-level logs and that the retailer can retrieve and produce them for regulators. Establish how that access works in practice, including who supplies the explanation when a complaint arrives.

The same review should cover how experiments are configured. Ask whether tests can operate at store or time level without assigning individual shoppers to higher-price conditions. Inspect the configuration process so the intended test unit matches the experiment that actually runs.

Finally, ask how consent and jurisdiction change the output. The vendor should explain how consent flags affect decisions and how Maryland, New Jersey and Connecticut receive different treatment, including group coverage, broader retailer coverage and private litigation exposure. A single national setting needs to satisfy the strictest standard applicable to the activity it governs.

Those answers determine which requirements the product can enforce and which need additional controls around it. A Maryland-compatible segment configuration may require changes in New Jersey, while Connecticut requires attention to a broader set of retailers. Procurement should make access to model inputs, output controls and retrievable logs part of the contract alongside commercial terms and marketing performance, with clear responsibility for supplying decision records when needed.

Final thoughts

Personalized grocery offers now require decisions that extend beyond marketing performance. When shopper data affects what someone pays or which discounts they can access, pricing, legal, technology and marketing leaders need shared accountability. A campaign’s conversion rate cannot establish whether its savings are credible, its allocation is fair or its treatment is permitted in each state.

Executives should make those questions part of campaign approval and vendor renewal. Require a justified reference price, a documented basis for offer eligibility and retrievable records showing how each decision was made. Give teams authority to stop campaigns when the evidence is incomplete, and assess complaints and repeat purchases alongside immediate revenue.

The state differences also make a single national configuration a choice that needs justification, rather than a default. Leaders must decide where to apply a common standard and where local rules require different controls, with clear ownership for maintaining both.

Personalization still has room to deliver value through loyalty benefits, retention offers and other permitted approaches. The durable business case rests on savings the retailer can substantiate and access it can defend. Those are operating requirements for maintaining customer trust, not simply records to assemble after a complaint.