What the Consumer Reports Instacart Pricing Test Found

What Consumer Reports Found in Its Instacart Pricing Test

Consumer Reports documented that shoppers could see different item prices for the same products from the same Instacart store during coordinated sessions. The investigation found variation. It did not independently prove that personal data or an estimate of each shopper's willingness to pay caused the assignment.

Instacart described the mechanism as randomized, short-term A/B testing by product category and store location. It denied using personal, demographic, or user-level behavioral information to set item prices. Fourteen days after the investigation was published, Instacart ended item-level price testing on its platform.

Those three facts belong together. The shoppers saw different prices. The platform supplied a disputed explanation for how groups were assigned. The platform then stopped the practice.

How the investigation worked

Consumer Reports published its investigation on December 9, 2025 with Groundwork Collaborative and More Perfect Union. The project recruited 437 volunteers.

In September 2025, volunteers joined four coordinated online sessions. Participants in each group used the same retailer and store location, shopped at the same time, and added an identical basket of 18 to 20 items from Safeway or Target. Another Safeway and Target test took place in person in Washington, D.C. A later November session examined other retailers.

The volunteers placed items in their carts and took screenshots. They did not complete purchases. Consumer Reports says roughly 200 screenshot sets without execution errors supported its central analysis.

That difference between recruited participants and analyzed records matters. "The study had 437 volunteers" is accurate. "All 437 produced the final dataset" is not.

Coordinating the product, store, and time removed several ordinary explanations. It did not give researchers direct access to the platform's assignment code, experiment logs, or data fields.

flowchart LR
    A["Matched shopping sessions"] --> B["Displayed prices recorded"]
    B --> C["Variation measured"]
    C --> D["Platform explanation requested"]
    D --> E["Assignment cause remains a separate evidence question"]

What the shoppers observed

Consumer Reports reported that every analyzed shopper appeared to be placed in an item-price experiment. About three-quarters of the checked products appeared at more than one price. The reported differences ranged from seven cents to $2.56 per item and reached 23 percent in one example.

One Seattle Safeway session gives the clearest basket view. On September 4, 2025, 39 volunteers added the same 20 items. The basket totals ranged from $114.34 to $123.93. The $9.59 gap was 8.4 percent of the lowest basket. Only 8 percent of that group received the lowest total.

The investigation also reported different displayed reference prices for certain discounted products even when shoppers received the same final sale price. That observation raises a separate transparency question. A price test can change the amount paid, the apparent size of a discount, or both.

These findings are meaningful because they establish simultaneous variation under coordinated conditions. They do not establish that every Instacart retailer, item, or customer experienced the same pattern.

What the $1,200 estimate means

Consumer Reports used the observed basket variation and an Instacart household-spending assumption to estimate a possible annual swing of about $1,200 for a household of four.

The figure was an extrapolation. Researchers did not follow a representative family for a year and measure $1,200 in added charges. The sessions used selected baskets during a limited observation period, and participants did not buy the items.

The estimate is useful as a scale illustration. Small differences can accumulate when they affect frequently purchased goods. It is not a verified annual loss for every family, a forecast for any named shopper, or proof that higher variants would recur at the same rate.

Instacart strongly disputed the calculation, arguing that Consumer Reports extended one atypical basket and short-term tests into a year-long claim. A careful account should show the estimate, its assumptions, and the objection rather than choosing the most dramatic shorthand.

What Instacart said the test was

Instacart published a detailed response on December 18, 2025. It accepted that shoppers were temporarily shown different item prices but rejected the terms dynamic pricing and surveillance pricing.

According to Instacart, retail partners used short-term randomized A/B tests to measure price sensitivity at a group level. Customers were assigned by product category and store location. Some saw a price above the pre-test amount, while others saw a lower amount.

The company said personal information, demographics, ZIP code, income, and individual shopping behavior did not determine item prices. It also said retail partners controlled base prices.

Target complicated the last point. Target told Consumer Reports it had no formal business relationship with Instacart. Instacart later said it used public Target prices as a starting point, added an amount for operating and technology costs, and had tested ways to offset those costs. The company said it ended the Target tests.

The public record therefore contains agreement on the displayed variation and disagreement about how the practice should be characterized. Instacart's explanation is the best public description of its internal assignment method, but it is still a counterparty statement rather than an independent audit.

Why the test was not proof of surveillance pricing

The FTC uses surveillance pricing to describe individualized prices or offers informed by personal data. Its January 2025 study found that pricing intermediaries can use location, browsing behavior, demographics, shopping history, device context, and other signals.

That market capability does not prove that the Instacart sessions used those inputs. Consumer Reports said its sample was not large or representative enough to test statistical relationships between demographics and assigned prices. Instacart denied that personal or behavioral data set the item prices.

To establish surveillance pricing in a specific experiment, an investigator would need evidence connecting a personal-data field or inference to the assignment and resulting offer. A screenshot proves what appeared. An assignment log, data dictionary, system specification, or auditable model record can help explain why it appeared.

[[How to Audit an Online Pricing Study]] turns that distinction into a reusable evidence method.

What changed after publication

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

The change did not eliminate every source of difference. Retailers can set different prices by store, choose online markups, and run promotions or loyalty offers. Weighted products, fees, fulfillment method, membership, and coupon eligibility can still change a total.

Instacart's July 2026 pricing principles restate the same-item commitment and say the company will not use personal information to set item prices, membership prices, or fees. The principles also preserve personalized discounts.

The New York attorney general sent Instacart a January 2026 information demand. The office warned that disclosures might not comply with New York's algorithmic-pricing law. That was an inquiry and warning, not a final finding that the tested prices used personal data.

The investigation's durable result

The investigation made a hidden experiment visible. It showed that a platform could place shoppers into price variants for essential goods without an obvious notice at the moment of choice. It also showed why output evidence and causal evidence must remain separate.

For shoppers, the practical lesson is to record the exact item, store, time, account state, promotion, and total before explaining a difference. [[How to Document Different Prices Online]] provides that packet.

For reporters and researchers, the lesson is to describe the strongest supported claim first: coordinated shoppers saw different item prices. The internal reason remains a second question.

For product teams, the lesson is that randomized assignment does not settle the fairness question. A customer can object to being placed in an undisclosed higher-price group even if no personal trait selected that group.

[[Algorithmic Grocery Pricing - What Shoppers Can and Cannot See]] distinguishes price testing, dynamic pricing, personalized pricing, and surveillance pricing. [[What Is a Loyalty Penalty]] explains why repeat purchase or customer tenure would need its own evidence before the E095 label could be applied to the test.

This explainer was developed from the Consumer Reports investigation and follow-up, Instacart's published responses and current pricing principles, FTC research, the New York attorney general's record, and the preserved E095 transcript. AI assistance was used for research organization, drafting, and validation. Publication remains unauthorized.