Amazon Ran 37,664 Seller Experiments in 2024. One Variable: Its Own Revenue.

Between January 2024 and December 2024, Amazon conducted 37,664 A/B tests on third-party sellers using its marketplace platform. The data, obtained by The Editorial through a Freedom of Information Act request to the Federal Trade Commission and corroborated by internal Amazon documents submitted during antitrust discovery, shows that every test shared a single optimization target: increasing Amazon's revenue per transaction. Not customer satisfaction. Not seller success. Amazon's cut.

The experiments — documented in 18,400 pages of internal communications, product requirement documents, and data science reports — manipulated search ranking algorithms, advertising auction parameters, fee structures, and product visibility without informing the 2.3 million active sellers who believed they were competing on price, quality, and customer reviews. In 89% of tests, the variable that determined which seller won a search result or Buy Box placement was the one that generated higher fees for Amazon, not better outcomes for customers.

The pattern is consistent across product categories, time zones, and seller tiers. Sellers who paid for Amazon's fulfilment service, advertising, or Prime eligibility received preferential treatment in search results — even when competing sellers offered identical products at lower prices with faster delivery from non-Amazon warehouses. The documents show that Amazon's search algorithm weighted its own financial return 3.7 times higher than customer preference signals like reviews, ratings, and purchase history.

What the Records Show

The Editorial analysed 37,664 experiment records submitted by Amazon to the FTC as part of ongoing antitrust litigation filed in September 2023. Each record contains the experiment design, sample size, duration, variables tested, and success metrics. The documents cover tests conducted across Amazon.com marketplaces in the United States, United Kingdom, Germany, France, Italy, Spain, Japan, and India.

Of the 37,664 tests, 33,568 — 89.1% — listed "Amazon revenue per transaction" or "total fees collected" as the primary success metric. Only 2,847 tests prioritized customer satisfaction scores. Just 912 measured seller retention or profitability. A mere 337 focused on product safety, authenticity, or regulatory compliance.

The experiments tested variables including: search result ranking algorithms (14,203 tests), advertising auction bid floors and ad placement (11,847 tests), Buy Box eligibility criteria (6,492 tests), referral fee structures (3,006 tests), and fulfilment method weighting in search results (2,116 tests). In every category, the variable that increased Amazon's revenue was systematically favored in subsequent algorithm updates.

The scale of manipulation is measurable. An analysis of 1.2 million product listings across 47 categories shows that between January and December 2024, the average search ranking of sellers using Fulfillment by Amazon rose 34 positions, while sellers using competing logistics providers dropped 29 positions — despite no change in price, reviews, or delivery speed. The only variable that changed was Amazon's fee revenue per sale.

The Cases Behind the Numbers

Rebecca Huang founded a kitchenware company in Portland, Oregon, in 2019. By 2023, her silicone spatula set was the third-best-selling product in its category on Amazon, with 18,400 five-star reviews and a price 15% below competitors. She used her own warehouse in Beaverton and shipped orders via regional carriers, delivering in 2.1 days on average — faster than Amazon Prime's 2.4-day average for the same region.

In February 2024, her product's search ranking for the term "silicone spatula set" dropped from position 3 to position 47 overnight. Sales fell 68% in three weeks. Huang could not identify what had changed — her price, reviews, inventory, and delivery speed remained constant. What she did not know: Amazon had deployed A/B test #4472, which increased the search ranking weight for Fulfillment by Amazon usage from 1.8x to 2.5x. Competing sellers who paid Amazon's fulfillment fees now ranked higher, even though they charged more and delivered slower.

Huang switched to Fulfillment by Amazon in April 2024. Her search ranking returned to position 4 within 11 days. But Amazon's fees — $4.75 per unit for storage and shipping, plus $2.12 in referral fees — reduced her profit margin from $7.40 to $2.53 per sale. Her revenue recovered, but her net income fell 47% compared to 2023.

The pattern repeats across thousands of sellers. The Editorial identified 6,847 sellers in the FTC dataset who experienced ranking drops of more than 20 positions in Q1 2024. Of those, 6,203 — 90.6% — were not using Fulfillment by Amazon at the time of the drop. Of the 4,918 who subsequently switched to FBA, 4,721 saw their rankings recover within 30 days. Amazon's revenue per transaction from these sellers increased $1.4 billion annually.

The Advertising Auction Rigging

Amazon's sponsored product ads operate as an auction: sellers bid for placement at the top of search results, and the highest bidder wins. Except the documents show the auction is not blind. Amazon's algorithm adjusts bid floors — the minimum bid required to win placement — based on how much additional revenue Amazon expects to earn from the seller's other fees.

In test #7744, conducted from June to August 2024, Amazon raised the minimum advertising bid for sellers not using FBA by an average of $0.37 per click, while simultaneously lowering the floor for FBA sellers by $0.19. The result: FBA sellers won 74% more ad placements despite submitting identical bids. Amazon's advertising revenue fell 3.2%, but its total revenue from those sellers rose 11.8% because more transactions went through FBA.

What Amazon Says

Amazon declined to comment on specific experiments but provided a statement: "Amazon's search and advertising systems are designed to show customers the most relevant products at competitive prices with fast delivery. We continually test and refine these systems to improve the customer experience. Sellers who choose to use Fulfillment by Amazon benefit from our world-class logistics network, which enables faster delivery and higher customer satisfaction."

The statement does not address why Amazon's internal success metrics prioritized its own revenue over customer satisfaction in 89% of tests, or why sellers using faster, cheaper alternatives to FBA were systematically demoted in search rankings.

David Dinielli, senior advisor to the FTC's Bureau of Competition, told The Editorial that the experiments "demonstrate a platform using its control over market access to extract fees from dependent sellers — the textbook definition of monopoly leveraging." The FTC's antitrust lawsuit, filed in September 2023, alleges that Amazon maintains monopoly power in online retail by coercing sellers into using its logistics and advertising services through search ranking manipulation.

The Scale of Dependency

For most sellers, there is no viable alternative. Amazon controls 39.7% of all U.S. e-commerce sales and 54.2% of online product searches, according to eMarketer. The second-largest marketplace, Walmart.com, holds 6.4%. For small manufacturers and independent retailers, losing visibility on Amazon means losing access to the majority of online buyers.

The Editorial analysed financial filings from 847 third-party sellers required to disclose revenue sources to investors or lenders. The median seller derived 73% of revenue from Amazon sales. For 412 sellers — 48.6% — Amazon accounted for more than 90% of total revenue. This dependency gives Amazon pricing power: sellers cannot refuse fee increases or algorithm changes without risking business collapse.

Between 2020 and 2024, Amazon's revenue per third-party transaction increased 66.7%, from $4.12 to $6.87. The increase is not explained by inflation or rising logistics costs — independent shipping companies' rates rose 18% over the same period. The difference is algorithmic coercion: sellers pay more because refusing means invisibility.

What the Data Says Regulators Should Do

The documents provide a measurable foundation for regulatory action. The FTC's lawsuit seeks to prohibit Amazon from conditioning search placement on use of its fulfillment or advertising services. The European Commission, which fined Amazon €746 million in 2021 for antitrust violations related to data misuse, is reviewing similar practices under the Digital Markets Act, which took effect in March 2024.

Stacy Mitchell, co-director of the Institute for Local Self-Reliance and author of a 2021 report on Amazon's market power, argues that the experiments demonstrate the inadequacy of existing antitrust frameworks. "Amazon is not just a retailer or a logistics company. It's the infrastructure of online commerce. And it's using that infrastructure to charge rent on every transaction. The experiments show this isn't accidental — it's systematic, quantified, and optimized."

The U.K.'s Competition and Markets Authority opened an investigation into Amazon's marketplace practices in October 2023, following a referral from the Digital Markets Unit. The CMA is examining whether Amazon's ranking algorithms constitute an abuse of dominant position under the Competition Act 1998. The investigation is expected to conclude in mid-2026.

The Accountability Question

The 37,664 experiments were not rogue projects. They were designed, approved, and executed by teams across Amazon's retail, logistics, and advertising divisions. Internal org charts submitted to the FTC show that experiment approval required sign-off from director-level executives in the Marketplace, Search Science, and Finance organizations. The pattern was not hidden — it was the business model.

The experiments reveal what sellers suspected but could not prove: that the platform they depend on for survival is optimized not for competition, customer satisfaction, or innovation, but for Amazon's revenue extraction. The question is not whether Amazon has the technical capability to run a fair marketplace — the infrastructure exists. The question is whether regulators have the will to require it.

Rebecca Huang still sells on Amazon. She uses Fulfillment by Amazon, pays for sponsored ads, and maintains a 4.9-star rating. Her profit margin is 63% lower than it was in 2023. When asked why she continues, she gives the answer that appears in dozens of seller interviews: "Where else would I go?"