The Loyalty Penalty: Venture Step Episode 95

The Loyalty Penalty: Dalton Anderson on the Instacart Pricing Investigation

Venture Step E095 begins with a question that feels ordinary until two people compare screens: what if shoppers at the same store, buying the same staple at the same time, do not see the same price?

Dalton Anderson recorded the episode after reading a Consumer Reports investigation into item-price experiments on Instacart. The report shifted his attention from familiar airline and hotel pricing to milk, bread, eggs, and other goods that households buy repeatedly.

His reaction was immediate and skeptical. If a platform could change the price without making the experiment visible, how could a shopper know whether a deal was real, why a neighbor paid less, or whether repeated purchases had become a signal to charge more?

The recording captured the right concern before the public record could answer every causal claim.

What triggered the episode

Consumer Reports, Groundwork Collaborative, and More Perfect Union coordinated Instacart shopping sessions with hundreds of volunteers in 2025. Participants used the same store, time window, and basket while recording the prices shown to them.

The published investigation reported that shoppers sometimes saw different item prices under those matched conditions. About three-quarters of checked products appeared at more than one price, and one item's reported spread reached 23 percent.

That finding drove E095. Dalton was not reacting to the possibility that grocery prices change across cities or stores. He was reacting to simultaneous variation inside one digital storefront.

flowchart LR
    A["Consumer Reports documents price variation"] --> B["E095 asks whether loyalty can become a penalty"]
    B --> C["E096 examines surveillance pricing and willingness to pay"]
    C --> D["E099 asks what grocery transparency should provide"]

The episode became the opening move in a broader Venture Step pricing cluster. E096 followed the data and policy questions. E099 returned to the consumer's practical need to compare grocery prices and understand the basket.

Why staple goods changed the moral question

Dalton accepts that a plane seat or hotel room can cost more when demand rises and capacity is limited. His objection was not to every changing price. It was to a system that might evaluate one shopper differently without making the rule visible.

Groceries sharpen that discomfort. Many purchases are routine, time-sensitive, and difficult to postpone. A shopper using delivery because of work, disability, distance, caregiving, or transportation constraints may have less room to abandon the cart.

The episode sometimes moved from that concern to an unsupported conclusion. It suggested that older people, disabled people, and shoppers in food deserts were being targeted. The investigation described Instacart as important to those communities, but its participant sample did not prove that those traits determined a higher price.

That correction does not erase the fairness question. It makes it more precise. A randomized test can still place someone into a higher-price group without notice. It simply is not the same claim as using disability, income, location, or purchase history to choose the group.

What E095 got right

The episode understood that a small item difference can matter when it recurs across a household basket. It also recognized that a displayed discount and a final price are two separate parts of an offer.

Consumer Reports reported examples in which shoppers saw different reference prices for discounted products even when the final sale price matched. That can change the apparent generosity of a deal without changing the amount paid.

E095 also saw that the old physical price tag had become a software decision. Digital retail systems can update an item, promotion, fee, rank, or message far more quickly than a clerk can replace a paper label. The operating cost of experimentation has fallen.

The durable question is therefore not whether software should be allowed near a price. It is what the shopper should be told when software assigns a treatment that changes the price or the meaning of a deal.

What later research corrected

Dalton described the test as AI-enabled willingness-to-pay pricing based on personal behavior. The investigation observed different outputs, but it did not independently inspect the full assignment system.

Instacart's December 2025 response said the assignments were randomized by product category and store location. The company denied using personal, demographic, or user-level behavioral information to set item prices.

The transcript also treats the $1,200 household figure as a typical annual cost. Consumer Reports produced that number by applying observed basket variation to a household-spending assumption. It was an estimate, not a year of measured charges for each family.

Eversight was not a separate company operating outside the platform. Instacart acquired 100 percent of Eversight in 2022 and integrated the technology into its pricing and promotion products.

Finally, a later $60 million FTC matter did not validate the price-test allegations. That case concerned delivery advertising, refunds, satisfaction guarantees, free trials, and Instacart+ enrollment. It was a separate dispute.

[[What Consumer Reports Found in Its Instacart Pricing Test]] reconstructs the method, observed results, estimate, response, and later developments in detail.

The system changed after the recording

On December 22, 2025, Instacart ended all item-price tests. The company said retailers could no longer use Eversight technology for item-price testing and that two shoppers using the same store location at the same time would see the same item prices.

Promotions, discounts, loyalty offers, retailer markups, and store-to-store differences remained. Instacart's July 2026 pricing principles repeat the commitment on item prices and say the company will not use personal information to set item prices, membership prices, or fees.

The change gives E095 a complete arc that the original recording could not have. The investigation documented a practice. The platform disputed the characterization. Public pressure and regulatory questions followed. The platform ended the item-price tests.

The loyalty penalty remains a testable question

E095 uses loyalty penalty to describe the fear that a repeat shopper could pay more because a system expects that person not to switch.

The established term comes from markets such as insurance, broadband, mortgages, and savings. Regulators have used it when longstanding customers pay more than comparable new customers because suppliers expect them to remain.

The Instacart investigation did not establish that purchase history or customer tenure caused the observed assignments. The episode title should therefore be read as a question and viewpoint, not as a causal finding.

[[What Is a Loyalty Penalty]] provides the definition and comparison needed to use the phrase carefully.

What the episode leaves behind

The most useful sentence to carry forward is not that every difference proves profiling. It is that a price difference deserves an explanation.

One screenshot can start the inquiry. A controlled comparison can establish variation. Repeated observations can show a pattern. The assignment rule, input data, and decision records are what turn that pattern into an explanation.

[[How to Document Different Prices Online]] helps shoppers preserve the first record. [[How to Audit an Online Pricing Study]] helps reporters, researchers, and product leaders decide how far a claim can go.

For the wider grocery context, continue to E099's [[Algorithmic Grocery Pricing - What Shoppers Can and Cannot See|Algorithmic Grocery Pricing: What Shoppers Can and Cannot See]]. For the personal-data question, continue to E096's [[Surveillance Pricing and the Invisible Negotiation]].

This episode story was developed from the immutable E095 transcript and current Consumer Reports, Instacart, FTC, and New York records. AI assistance was used for research organization, drafting, and validation. The original recording remains unchanged, and the corrections above govern the public version.