Cut the Commission, Feed the Competitor: The Take-Rate Trap

Based on the research ofMa, "Pricing, Matching, and Bundling: an Equilibrium Analysis of Online Platforms," arXiv preprint, 2026

A commission cut reads as a clean loss on a spreadsheet: revenue down, everything else held constant. Gary Qiurui Ma's equilibrium analysis of platform design argues that "everything else held constant" is the mistake. Pricing, matching, and bundling are not independent settings a manager can tune one at a time; they are complementary levers in a single equilibrium, and participants on every side of the market watch what the platform does with each one and adjust who joins, what they list, and how hard they compete. Cut the take rate and you change who can profitably sell on your marketplace. Change who can profitably sell and you change what your search ranking is worth. Change what ranking is worth and you change whether it is more profitable to source the product yourself and sell it as a first-party bundle. None of this shows up if you model pricing alone.

One Complaint, Three Levers

The clearest real-world proof of this coupling is a single antitrust complaint. When the Federal Trade Commission and seventeen state attorneys general sued Amazon in September 2023, they did not describe three separate abuses; they described one connected mechanism spanning all three levers Ma's thesis analyzes.

The pricing piece is the most familiar: the complaint alleges that Amazon's combined seller fees, monthly charges, fulfillment costs, and now near-mandatory advertising spend "force many sellers to pay close to 50% of their total revenues to Amazon." But the FTC did not stop at price. It described how Amazon enforces a pricing rule through the matching lever rather than a price penalty: if a seller lists the same item cheaper elsewhere, the complaint alleges, Amazon can "bury discounting sellers so far down in Amazon's search results that they become effectively invisible." That is a commission-style punishment delivered entirely through ranking. And the complaint explains why ranking matters at all in blunt terms, accusing Amazon of "biasing Amazon's search results to preference Amazon's own products over ones that Amazon knows are of better quality." Pricing enforcement running through matching, and matching tilted toward the platform's own goods, are not two separate line items in a legal filing. They are one strategy, and separating them analytically is exactly the error Ma's equilibrium framework corrects.

Biasing Amazon's search results to preference Amazon's own products over ones that Amazon knows are of better quality.

The Product Built From Its Own Competitors' Data

Bundling is the least intuitive of the three levers, because it does not look like a pricing or ranking decision at all. It looks like a merchandising choice: the platform decides to source and sell a product itself. But Ma's framework treats bundling as a lever precisely because it interacts with the other two. A platform that can see every seller's sales data holds information no ordinary seller has, and a platform that also controls the ranking deciding whether its own bundle gets seen holds a second advantage no ordinary seller has either.

That combination is exactly what a Wall Street Journal investigation, reported by CNBC in April 2020, found inside Amazon. Reporters identified executives who had built workarounds, internally nicknamed "going over the fence," to access non-public, seller-specific sales data used to decide which private-label products Amazon should launch to compete with those same sellers. Amazon's own associate general counsel had told the House Judiciary antitrust subcommittee under oath in July 2019 that the company did not use individual seller data to develop competing products or to favor its own listings in search. The Journal's reporting directly contradicted that testimony. Amazon disputed the specific claims but confirmed it had opened an internal investigation; the underlying practice, a private-label arm sourcing decisions from the marketplace it also operates, is not disputed at all. It is the business model.

This is bundling doing exactly what Ma's thesis says it does: it "shapes the structure of supply when the platform itself acts as a market participant." A platform that bundles is no longer only a referee setting rules for other players; it is also a player, competing for the same sale it referees, using data and shelf placement no other competitor in that game can access. The commission a third-party seller pays and the ranking deciding whether their listing or Amazon's own is seen are no longer neutral background for that seller. They are terms set, in part, by their own landlord's rival product line.

Regulators Are Chasing the Coupling, Not the Levers

Antitrust authorities have converged on the same reading independently, which is itself evidence the coupling is real rather than a theoretical construct. The European Union's Digital Markets Act does not treat platform commissions and search rankings as separate problems. On March 25, 2024, the same day it opened formal non-compliance investigations against Alphabet, Apple, and Meta, the European Commission also launched a fact-finding step to clarify whether Amazon was "preferencing its own brand products on the Amazon Store," specifically under Article 6(5) of the DMA, the provision barring a gatekeeper from treating its own products more favorably than equivalent third-party offerings in ranking and related access. The formal proceedings that same day targeted Alphabet's steering rules in Google Play and self-preferencing in Google Search, Apple's steering rules in the App Store and its browser-choice screen, and Meta's "pay or consent" model. One release, five separate lines of inquiry, one shared theory: a platform's ranking decisions and its own commercial interests cannot be evaluated apart from each other, or apart from the fees it charges the participants that ranking affects.

Apple's own commission history shows the limits of moving only the pricing lever. In November 2020, Apple announced the App Store Small Business Program, halving its commission from 30 percent to 15 percent for developers earning up to one million dollars a year, framed by Tim Cook as help for small developers, and backed by Apple's citation of an Analysis Group study finding its commission structure "in the mainstream" for app-distribution platforms. That is a real, generous pricing move. It is also, on its own, a move that leaves matching and bundling untouched: it does nothing about how the App Store's own search and category placement treats Apple's apps and services relative to the small developers who just got a fee cut, which is exactly the gap the DMA's steering and self-preferencing investigations were opened to examine four years later. A price concession did not settle the equilibrium; it just moved the argument to the other two levers.

What Happens When You Pull Just One Lever

The cleanest natural experiment for Ma's core claim did not come from a tech giant at all. It came from city governments capping food-delivery commissions during the pandemic, a pure, single-lever pricing intervention: regulators told DoorDash and its rivals they could not charge restaurants more than a fixed percentage per order.

DoorDash's own account of what happened next, published on its company blog and reported by Restaurant Business in 2022, demonstrates the cross-lever response the paper's equilibrium logic predicts. Facing caps in more than seventy jurisdictions, DoorDash did not simply absorb the lost commission revenue. It reopened the pricing lever from a different angle, adding new consumer-facing fees of one to two dollars per order in capped markets to recover the margin regulators had removed from the restaurant side. The result was not neutral: DoorDash reported order volumes falling roughly 7 percent in Philadelphia, 4 percent in St. Louis, and 4 percent in Westchester County, New York, the month those fees took effect, attributing the declines directly to the price increase. "We have seen a tangible impact of the basic economic rule," the company wrote. "When prices go up, demand goes down."

Notice what the regulators accomplished and what they did not. They successfully moved the pricing lever, capping the commission restaurants paid. They had no mechanism to prevent the platform from reopening a different pricing channel aimed at a different side of the market, and none at all touching matching or bundling. The commission cap did not settle into the equilibrium regulators pictured; it shifted where on the platform the price showed up, and depressed the transaction volume the policy was meant to protect.

That is the pattern connecting a food-delivery fee cap to an antitrust complaint to a continental regulatory regime: pricing, matching, and bundling are one system with three visible controls, and every real attempt to move just one of them, whether by a platform trying to look generous or a government trying to protect a constituency, ends up revealing how much of the answer was hiding in the other two.