The Keeper of the Fair

By the late Middle Ages, a merchant shipping goods down the Rhine could expect to stop again and again to pay. Castles stood on the cliffs above the river and many existed for little reason except to collect. The lords who held them had built nothing that helped the cargo move. They had simply positioned themselves where the boats had to pass, stretched chains or sent armed men across the water, and charged for the privilege of continuing. Some of those tolls were legitimate in their time. Many were not much more than extraction with a coat of arms. The ruins are lovely now. The trade they taxed was slower and poorer for them.

A few hundred kilometers to the west, something different was happening. Through the twelfth and thirteenth centuries, the counts of Champagne hosted a cycle of great fairs in Troyes, Provins, Lagny and Bar-sur-Aube. Merchants came from Italy, Flanders, Germany and beyond, and they paid to be there. The counts charged fees and took their share. But in exchange they offered things a merchant could not easily provide for himself: safe conduct on the roads, officials who enforced contracts and settled disputes on the spot, standard weights and measures, money changers and a predictable calendar that made it worth traveling hundreds of miles. Out of that institutional work came some of the early machinery of European credit and finance.

There have always been two kinds of toll. One is payment for passage. The other is payment for something. Every marketplace eventually has to decide which kind it is collecting and sooner or later its participants decide for it.

There have always been two kinds of toll. One is payment for passage. The other is payment for something.

What are we paying for?

Every platform business eventually hears the same question from the people who depend on it: what, exactly, are we paying for?

For most of the last fifteen years the question was easy to postpone. A marketplace, an app store, a payments network or a software ecosystem grows by being useful to two groups at once: buyers and sellers, users and developers, riders and drivers. Economists call these two-sided markets. The fee usually arrives late, after both sides have committed and built their habits and businesses around the platform. By the time anyone asks seriously, the platform has scale and scale usually ends the conversation.

That assumption is no longer safe, and the clearest evidence is in the courts. In the long dispute between Epic Games and Apple, a federal judge in California found Apple in contempt in April 2025. She had ordered Apple to allow developers to link out to other ways of paying and Apple had responded by charging a 27 percent commission on those outside purchases. In December 2025 the Ninth Circuit upheld the contempt finding but rejected a flat ban on any commission at all, sending the matter back so the lower court could fashion something properly tailored. On June 30, 2026, the Supreme Court agreed to hear Apple’s appeal. Google, facing related claims from Epic, settled and cut its Play Store commissions.

Whatever the courts ultimately decide, the question being argued underneath the legal doctrine is the one the Rhine merchants and the Champagne merchants could have answered without lawyers. What does the platform contribute to each transaction and is the price connected to that contribution?

The bazaar and the cost of not knowing

The anthropologist Clifford Geertz spent time in the 1960s and 1970s in the bazaar of Sefrou, a market town in Morocco, and wrote one of the most useful essays I know about how markets actually work. In the bazaar, he observed, information is “poor, scarce, maldistributed, inefficiently communicated and intensely valued.” The central problem for any buyer or seller is not the price. It is finding out who is reliable, what the goods are really worth and whether the person across from you will be there tomorrow. Search is the main cost of doing business.

The bazaar’s answer was what Geertz called clientelization. Buyers and sellers form lasting, repeated relationships with particular partners and trade mostly inside them. You go back to the same cloth merchant not because his prices are the lowest but because you have learned you can trust his measurements and he has learned you pay. The relationship is the infrastructure.

I find this a clarifying lens on modern platforms, because the best of them are machines for reducing the cost of not knowing. Reviews that mean something, verified identities, dispute resolution, fraud protection, payment guarantees, the assurance that a buyer or a counterparty will actually be there: these are the descendants of the wardens of the Champagne fairs. They solve the bazaar’s problem at a scale Sefrou could never imagine. And a platform that genuinely solves it has earned its fee in the oldest sense.

Early in my career I worked on the trust and safety machinery of online marketplaces, including programs for handling counterfeit goods, fraud and abuse on some of the largest platforms of that era. It was unglamorous work. It was also, I came to understand, a large part of what those platforms were actually selling. The listings were the visible product. The confidence that a stranger would not cheat you was the real one.

Monetization is a design choice

Founders often treat monetization as a pricing exercise, something to settle after product-market fit. In practice, the way a platform charges shapes how everyone on it behaves, which makes it a design decision from the very beginning.

A transaction fee aligns the platform with volume, but it also gives buyers and sellers a reason to meet on the platform and complete the next deal somewhere else. A subscription is predictable, but it separates what participants pay from what they get and that gap becomes visible in a downturn. Advertising keeps access free while introducing a quiet tension between what users want to see and what the platform is paid to show them. Financial services layered on top, such as payments, lending or holding balances, can become the most profitable layer of all and they bring obligations that change what kind of company the platform has become.

None of these models is wrong. Each one sends a signal about whose interests the platform is organized around and participants read that signal quickly, often more quickly than the platform’s own leadership does.

Durable fees pay for something scarce

The platforms that hold their pricing over time tend to charge for things participants cannot easily assemble on their own. Trust is the most common. Liquidity is another, the assurance that the other side of the market will actually show up. Workflow and distribution round out the list.

When a fee is justified by trust and liquidity, people pay it with little complaint, because the alternative is doing that work themselves. When a fee is justified mainly by control, because the platform is the only door into a market, it becomes exposed on two fronts at once. Competitors can undercut it and regulators and courts can decide the price no longer reflects any service.

That is the Rhine castle problem. A toll for passage holds only as long as there is no other river.

The writer Lewis Hyde made a related argument about art in The Gift, published in 1983. He distinguished between gift economies, where things circulate and create bonds, and commodity exchange, where transactions are complete and leave no connection behind. Many of the platforms we now depend on began life feeling more like the first: communities of creators, early adopters and enthusiasts, with the business model deferred. When the fee finally revealed itself, people did not only recalculate their costs. They re-examined the relationship. The sense of betrayal that accompanies a platform’s late turn toward extraction is not irrational. It is what people feel when a gift relationship is revealed, after the fact, to have been a commodity relationship all along.

Leakage is information

Most platform operators treat disintermediation as a problem to be policed. Participants who meet on the platform and then take their relationship elsewhere are seen as lost revenue and a rules violation.

Geertz would have recognized it immediately as clientelization. Once two parties trust each other, they no longer need the institution that introduced them. A marketplace for recurring, high-value services will leak once a client and a provider have worked together a few times, because the platform’s main contribution was the introduction, and the introduction has been made.

A more useful reading is that leakage is data. It shows precisely where participants believe the platform stops adding value. The strongest response is rarely a stricter rule. It is a reason to stay: payment protection, scheduling, records, financing, insurance or anything else that keeps the platform useful after the first transaction. The Champagne fairs did not keep merchants coming back by forbidding them to trade elsewhere. They kept them by being the best place to trade.

A view from where the rules get tested

For about twenty-four years I worked as a litigator and counselor in technology, cybersecurity, entertainment and media. A lot of that work sat at the points where the rules governing technology businesses were being tested under pressure, often before the law had settled. What stays with me from those years is how often a dispute about a fee was really a dispute about a story and about whether the people paying could tell it.

The fee is rarely the real issue. The issue is whether the people paying it can explain what they get for it. When they can, pricing is a negotiation. When they can’t, it tends to become a dispute sooner or later.

That distinction matters on both sides of the relationship. For the platform, it is a test of whether its pricing will last. For a company building on someone else’s platform, it is a way to judge how much risk sits inside a distribution channel that looks free today.

What to settle before the question arrives

Operators who want pricing that lasts can do most of the work in advance.

Map what the platform contributes to each side separately, since buyers and sellers rarely value the same things. Price the portion of that contribution you could defend to a skeptical customer, a competitor or a court. Watch where participants leave and treat each departure as a product brief. Be cautious about any advantage that rests mainly on switching costs, because switching costs are exactly what courts and regulators have become comfortable examining.

For companies building on a platform, the discipline runs in reverse. Know which parts of the business depend on a single gatekeeper’s pricing and have a plan for the day that price changes.

When the buyer is a machine

There is a new chapter of this story beginning and it may change the bargain again.

For all of history, the people choosing where to trade have been people. They walked into the bazaar, traveled to the fair, scrolled the app store. Much of what platforms have built, from storefront design to recommendation engines to the careful placement of the “buy” button, is aimed at human attention and human habit. Software agents are starting to do some of that choosing for us: comparing offers, booking travel, reordering supplies, negotiating within limits we set. An agent does not get tired of searching. It does not feel the pull of a familiar interface. It does not stay out of loyalty.

If that shift continues, a great deal of what platforms currently charge for will be tested. The cost of search, the bazaar’s central problem, falls sharply when a machine can query a hundred sellers in a second. Fees that depend on being the place people habitually start will look more and more like Rhine tolls. Fees that pay for things an agent cannot supply for itself (verified identity, real recourse when something goes wrong, guaranteed settlement, accountability that a person can call on) may become more valuable, not less. The keepers of the fair will have new kinds of customers and those customers will be very good at reading the fee schedule.

I don’t think anyone knows yet how this settles. But it sharpens the old question rather than replacing it. When the buyer is a machine acting for a person, the platform still has to answer what it contributes. It just has to answer faster and to an audience that cannot be charmed.

The keeper of the fair

I think the medieval contrast holds up because it is not really about economics. It is about the difference between a relationship and a position.

The lord in the Rhine castle had a position. He stood where others had to pass and he collected because he could. The count of Champagne had a relationship. He collected because merchants chose to come and they chose to come because he made their lives safer, their contracts more reliable and their journeys worth the trouble. One built ruins. The other helped build the commercial world that followed.

Every platform, and in a sense every intermediary, every middleman, every institution that stands between people who want to trade, faces the same choice. It can be a toll on a river or a keeper of a fair. The platforms that come through the present period in good shape will not necessarily be the ones with the lowest fees. They will be the ones whose customers can say, without hesitating, what the fee buys.