The direct deal promises that the difference can remain with the parties who actually want the exchange. The maker receives more, the user pays less and the person between them disappears. It is an attractive picture because the middleman's invoice is visible while much of his work is hidden in the completed transaction. Once the invoice is removed, a buyer still has to find the good, judge the offer, arrange payment and receive what was promised. Some of those tasks can become almost effortless. Others move to a different account. The saving depends on which change has occurred.
A package sent from the maker can still pass through a payment service, a delivery network and a system that supplied the customer. The goods may follow a shorter route while their sale depends on several specialized organizations. Conversely, a producer can do the selling itself, and a customer can perform a search without buying a search service. It would be a mistake to call every remaining task another middleman. A function is an activity; an intermediary is an arrangement for having someone perform it. Disintermediation can remove the arrangement, reorganize the activity or make part of the activity unnecessary.
Sarkar, Butler and Steinfield approached the early electronic marketplace through precisely this distinction between a physical channel and the services around it. A technology that makes direct contact cheaper can also make it cheaper to coordinate specialized providers. Their 1998 account therefore offered reasons for new intermediaries to appear alongside opportunities for direct dealing. The important comparison is among the costs of completing the exchange. The technology's effect on the number of firms is an outcome to investigate, not a definition of whether the exchange has become better.
Imagine a maker receiving one hundred for a good. The intermediary charges the customer a further twenty-four for the services needed to obtain it. Suppose those services use fourteen in economic resources, leaving ten to the intermediary beyond those costs. A new route offers the same good and service quality for the maker's hundred plus a fee of nine, but requires the customer to perform work worth six. Assume that the new provider uses seven in resources and keeps two. The visible fee has fallen by fifteen; the customer's full cost has fallen by nine, from one hundred and twenty-four to one hundred and fifteen.
The resource saving is smaller still. The old arrangement used fourteen between maker and user; the new one uses seven at the provider and six at the customer, or thirteen. One has been saved in total. The customer gains nine, while the intermediary's surplus falls from ten to two, a reduction of eight. Those changes add to the one-unit resource gain. The maker's receipt, the quantity traded and the value of the finished good are held fixed throughout. Cutting a fee has both improved this transaction and changed who receives its benefit. Neither effect has to be disguised as the other.
If a competing provider instead performed the same work at the same resource cost of fourteen and charged fourteen, the customer would gain ten and the incumbent would lose its ten of surplus. At unchanged output, the movement would be a transfer. It could matter greatly to those making the exchange without saving ten in resources. If an innovation then reduced the necessary work from fourteen to four, a further ten could be a real saving. These are separate achievements. The first concerns the terms on which access is sold; the second concerns what access requires.
The work can also move inside the producer's organization. A manufacturer may find that its knowledge of the product makes advice, returns and repair cheaper to provide directly. It may instead discover that processing thousands of small orders distracts from production and costs more than using a distributor. There is no contradiction in a firm bypassing a wholesaler while hiring a payment processor or a specialist carrier. The relevant boundary can move separately for each task. A shorter list of firms can conceal more expensive internal work; a longer list can coordinate specialties that make the whole undertaking cheaper.
Digital verification offers a particularly sharp version of the distinction. A shared ledger can make a recorded transaction easier to check without asking one organization to certify its own account. But verifying an entry is different from establishing that the physical event described by it occurred. A ledger that accurately preserves a claim about a shipment does not itself inspect the shipment. Catalini and Gans distinguish the low cost of verifying digital state from the cost of maintaining a reliable connection to events outside the system. That connection may require inspection, devices, institutions or a trusted provider even when the record itself is distributed.
A new intermediary can accordingly grow out of a successful bypass. Once it brings together enough users, a producer may find access to those users difficult to replace. A search facility can become the place at which the search begins; a system assembled to make trading easier can become the system within which a trader must remain. This possibility does not establish that greater size always means a greater toll. Scale can lower costs and improve a match. The question is whether those gains remain available through competing routes, and whether the new provider can restrict an alternative while continuing to benefit from the people already gathered around it.
The old intermediary's disappearance is therefore incomplete evidence about its past. A new technology can render a once-useful task unnecessary or allow another organization to perform it more cheaply. Neither change proves that the earlier charge was entirely extraction. Its survival is equally incomplete evidence in its favor. An organization can survive because it remains useful, because customers cannot coordinate a move or because it can obstruct access to the alternative. Sarkar and his coauthors placed power and established relationships alongside transaction costs in their analysis. The comparison has to follow the conditions of exit as well as the technical ability to send an order.
That is why the dream of an exchange with no institutions is less useful than the prospect of an exchange with better ones. Direct dealing still relies on terms, records and ways of responding when a promise fails. Those arrangements can be simple, shared or largely automated; they can also become another source of dependence. Coase's insistence that an actual market needs rules directs attention to this continuing work. Removing a firm can be a genuine improvement. The improvement is established by what the parties can now do, what it costs them and what they can do if the new arrangement ceases to serve them.
A bypass succeeds when the parties gain a better way through, not merely when the invoice changes its name.
The direct deal can free the parties by reducing the work between them, by giving them an alternative or by returning a larger part of the gain to their hands. It can also place them behind a new gate. These possibilities give the original question its practical force: follow the work that survives, the costs that move and the choices that remain. The final plate asks what should happen when the best route is also a route one owner can close, and when another passage would be too costly to build. The argument remains open.