The dealer, the bank, and the arbitrageur all plied their trades within markets that existed before them and would exist without them; they stood in a marketplace, but they did not build it. The middleman of this plate builds the marketplace itself. The platform — the exchange, the marketplace, the app, the network — does not find a market and insert itself; it creates the venue in which a market can happen, gathers the two sides who will trade there, and then stands in the middle of the thing it has made, charging each side for the meeting it alone can host. Its price has a name of its own: the take rate, the cut it skims from every transaction that passes through it, the percentage of each sale or fare or order that the venue keeps for itself. The app store's portion of every purchase, the card network's slice of every swipe, the marketplace's fee on every sale, the auction house's premium and commission, the rideshare platform's share of every ride — all are take rates, the price of standing in the middle of a market one has built. And the take rate is, of all the prices in this volume, the one that lies most exactly upon the seam the volume has been tracing, for it is at once the clearest wage of a genuine service and the heaviest toll of an unrivalled position, and the two are bound together in it more tightly than anywhere else.
To see why, one must see what kind of market a platform is, and the decisive analysis is that of Rochet and Tirole, who named it the two-sided market. The platform serves not one kind of customer but two, who need each other: buyers and sellers, riders and drivers, merchants and cardholders, readers and advertisers. And the worth of the platform to each side depends on how many of the other side it has gathered — a marketplace is valuable to a buyer in proportion to the sellers on it, and to a seller in proportion to the buyers. This mutual dependence, which the economists call a network effect or a cross-side externality, is the platform's defining feature and its central difficulty. It makes the launching of a platform a chicken-and-egg ordeal, for neither side will come before the other: the sellers will not stock an empty marketplace, and the buyers will not visit one without sellers. And it makes the pricing of a platform a subtle, two-handed affair, for the platform must set a price to each side with an eye to its effect on the other, and will often charge the two sides very unequally — courting one side with low prices or none at all, in order to make the platform attractive to the side it means to charge. The structure of the take, and not merely its size, is the platform's chief instrument; whom it subsidises and whom it bills is the art of getting a two-sided market off the ground at all.
The network effect does something else, beyond complicating the launch and the pricing: it pushes the whole market toward a single winner. For if each side is drawn to the platform that has the most of the other side, then the platform that pulls ahead becomes more attractive precisely by being ahead, and its lead compounds — the buyers flock to it because the sellers are there, and the sellers because the buyers are, and each arrival makes the next more likely, until the market tips and one platform holds nearly all of both sides. This is the increasing-returns dynamic that Brian Arthur traced, the lock-in by which an early advantage becomes a permanent one, and it explains why two-sided markets so often end not in a field of competitors but in a single dominant venue. And the dominant venue, once established, is something more than large: it is indispensable. Each side is there because the other is; neither will leave, because there is nowhere else that both are; and no challenger can pry either side loose, because it cannot offer the one side the other until it has gathered both, and it cannot gather both because they are already settled on the incumbent. The network effect that drew the two sides together has become a wall around them, and the platform within the wall is no longer merely a place to trade but the place one must trade, the venue there is no leaving because everyone else is in it.
Each side comes because the other is there; once both are gathered, neither will leave, for there is nowhere else that both are — and the venue that holds them both can charge them both for the holding.
Now, none of this would be troubling if the platform were a mere parasite on the market it hosts, but it is not; it does real and valuable work, and a large part of the take rate is the honest wage of that work. The platform builds and maintains the venue — the technology, the infrastructure, the marketplace itself, which does not run itself and is not cheap to run. It solves, at scale and for everyone at once, the problem of search that an earlier plate set out, bringing the two sides together and sparing each the labour of finding the other. It manufactures trust between strangers who would otherwise have no reason to deal: it rates and reviews them, it holds their money in escrow, it adjudicates their disputes, it stands behind the transaction so that a buyer in one country will hand his money to a seller in another whom he will never meet. It handles payment and guards against fraud; it manages, often, the logistics of fulfilment and delivery. These are substantial services, and they are services the two sides genuinely could not easily provide for themselves — the very services whose absence, in a market without a platform, leaves buyers and sellers to find and vet and trust and pay one another at ruinous cost. For all of this the take rate is a fair price, the wage of a venue genuinely built and genuinely useful, and a platform that charged for this and no more would be earning its cut as honestly as the dealer earns his for immediacy and the bank earns its for monitoring.
But the platform that has tipped its market and walled it off is in a position to charge for something quite other than its service: it can charge for its indispensability. Once both sides are locked within the network, with nowhere else to go that has the other side, the platform can raise its take rate — widen its cut, squeeze the sellers, dun the buyers — and lose neither side, for neither side has an alternative to lose it to. The take rate then ceases to track the cost or the value of the venue's services and begins to track, instead, the simple fact that the venue cannot be done without. This is the App Store's contested thirty percent, levied on developers who have no other way to reach the platform's users; it is the interchange fee that merchants must pay because they cannot refuse the dominant card; it is the marketplace seller's fee, ratcheting upward year by year, paid by sellers whose customers are all on the one platform and who therefore cannot leave it. In each case the take rate has become, in part, a toll — a charge not for a service rendered but for a passage that cannot be avoided, the price the platform sets upon its own unleavability. It is the same toll the volume has met before, in the dealer who quotes wide because no rival quotes narrow and the bank that lends dear because the borrower has nowhere else to turn; but the platform's toll is the heaviest of all, for the platform's chokepoint is not a momentary advantage but the whole market, walled and held.
An indispensable venue can charge for its indispensability; the take rate of a platform no one can leave is, in part, a levy on the captive, priced not by the cost of the service but by the want of an alternative.
Here, then, is what makes the take rate the volume's sharpest case, and not merely another instance of its theme: in the platform, the wage and the toll cannot be separated, because they spring from the very same thing. The platform's value lies in its network — in the gathering of many buyers and many sellers into a single place — and the larger that network, the more valuable the platform genuinely is, for a bigger network means more matches, more choice, deeper liquidity, a better market for everyone in it. But the larger that network, the more completely it locks both sides in, and the more unleavable the platform becomes, and the heavier the toll it can exact. The network is at once the service and the source of the power; the thing that earns the wage is the thing that confers the toll; and there is no way, even in principle, to have the one without the other, for a network large enough to be maximally useful is a network large enough to be inescapable. The dealer's spread folded the honest and the extracted into a single number that would not disclose their proportions; the platform folds the honest and the extracted into a single asset, the network, that is genuinely both at once — valuable and entrapping in the same breath, and by the same fact. One cannot point to the part of the network that serves and the part that ensnares, for they are not two parts but one.
And so the take rate poses the volume's question in its most intractable form. A high take rate may mean that the platform's network is enormously valuable — that it furnishes matches and trust and reach so great that a large cut is a fair share of a large benefit — or it may mean that the platform's network is enormously entrenched, that the cut is a levy wrung from a captive who cannot escape it. It may be the wage of a service worth every point of it, or the toll of a position that would command the same points if the service behind it were half as good. And the number itself will not say which, for the two interpretations describe the identical fee, levied through the identical network, on the identical locked-in sides. The whole apparatus of competition policy, which labours over platforms more than over any other business of our age, is an attempt to answer a question that the take rate alone cannot — to judge, behind the number, whether the platform's cut reflects the value it creates or the exit it forecloses, and whether a venue that has become the market should be free to price its own indispensability or restrained from doing so. The verdict of this plate is only that the question is real and the answer is not given in the price: the take rate is, demonstrably and simultaneously, the wage of a network genuinely built and useful and the toll of a network that cannot be left or rivalled, earned and extracted in the same percentage, by the same fact of the same gathered crowd.
The platform is therefore the middleman of this volume raised to his highest power — not standing in a market but owning one, not charging for a service within a venue but charging for the venue itself, and pricing, beneath whatever service he renders, the plain fact that he cannot be done without. His take rate is the seam of the whole volume drawn tight: the wage and the toll, the earned and the extracted, fused in a single cut by the single network that is at once his service and his sway. What that sway amounts to, and how far the law should reach to check it — whether the power a network confers upon its owner is a fair prize for having built the better venue or a dominance that ought to be curbed however it was won — is the matter of the volume on power, to which the bridge from this plate runs, and where the rents of position are weighed apart from the services that position may also perform. The next plate, though, stays with the toll and strips away the service entirely, to look at the middleman in his barest and most suspected form: the one whose cut reflects no service at all, but only a chokepoint seized and held — the rentier of the passage, in whom the volume's oldest charge, of the parasite who merely levies, finds at last its kernel of truth. As the table says of every price, first to last: the argument remains open.