Platform Businesses: Why Two-Sided Markets Behave Differently

A platform business does not make a thing and sell it. The firm sets the terms on which two separate groups meet, and earns a share of what passes between them. The firm never takes title to the goods, so what flows through it and what it keeps are different numbers. Each side is worth more to the other as the other side grows. The mutual dependency makes a platform hard to start.

Everything strange about a platform comes from one structural fact, and it is worth saying plainly before anything else: the firm is not a party to the trade. The firm is the place the trade happens and the author of the rules under which it happens. Once that fact is held, the odd revenue figures stop being odd, the difficulty of the first year stops being a puzzle, and the reason two participants can read the same growth in opposite ways becomes obvious. All four follow from that one fact, and none of them needs anything beyond revenue and margin to follow.

What actually makes a market two-sided?

The word is used far too loosely, and a strict definition rules out most of what it gets applied to. Three things must all be true at once, and if any one of them is missing the subject is an ordinary business with a lot of customers. A market is two-sided when two distinct groups each need the other, and a firm writes the rules under which the two of them come together. Two groups. Mutual need. A firm writing the rules. Miss one and the label does not apply.

Consider the weekly vegetable market on a municipal ground. The municipality does not grow a single tomato. The municipality rents out pitches, decides who is allowed to set up, fixes the hours, and settles arguments about weights. Farmers come on Tuesday because shoppers come on Tuesday. Shoppers come on Tuesday because farmers come on Tuesday. Neither group would come for an empty ground. The municipality grows nothing and sells nothing, and the municipality is what makes Tuesday work. The Tuesday market is a two-sided market, and markets like it existed long before anybody built one on a phone.

Now hold that up against the vegetable shop on the corner. The shop has hundreds of customers, and it buys from dozens of farmers. The shop looks superficially similar and is a completely different animal. The shop buys the tomatoes, takes them into its own stock, and sells them on its own account. The farmers are its suppliers, not a side of a market. The customers are its buyers, not a side of a market. There is one trade, repeated, with the shop on one end of every single one. The two never meet, and the firm is a party to every trade rather than the author of the terms, so a business with many customers on one side and many suppliers on the other is not two-sided.

Setu Bazaar, an invented marketplace, works like this. A buyer opens it, finds a listingOne item put up for sale by one merchant, with its price, its description and its terms attached. The listing is what a buyer searches and what a marketplace publishes; it is not stock the marketplace holds. from a merchant in another town, and pays. The merchant packs the parcel and sends it. Setu Bazaar never sees the parcel. Setu Bazaar let the two find each other and stood behind the terms of the meeting.

Now apply the same three questions to Anjani Stationers Private Limited, an invented manufacturer selling paper products to a few dozen business buyers. Does it have two distinct groups that need each other? No. Anjani has customers on one side and a paper mill supplying it on the other, and its customers have no need whatsoever of its mill. Does it set terms on which two groups meet? No. Anjani sets its own prices for its own goods, and setting its own prices is a different act. Does it avoid taking title? No. Every ream is bought as pulp, made into paper, held as stock and sold on Anjani's own account. Anjani fails all three. A test that only ever passes teaches nothing, so the useful thing to do with the three-part test is to run it on a business that fails it and watch where the failure lands.

Why is what passes through a marketplace not its revenue?

Here is where most of the confusion about platforms lives, and it is an arithmetic point rather than a subtle one. Because Setu Bazaar never takes title to the goods, it never buys them and never sells them. Setu Bazaar therefore has nothing to report as the sale of goods. The fee it charged for running the meeting place is the only thing it has to report. The value flowing through a platform and the amount a platform keeps are two different numbers, and they are routinely confused, usually in the direction that flatters the platform.

Look at the consequences one at a time. Each of them shows up somewhere different in the accounts. The platform carries no inventoryGoods a business has bought or made and still holds, waiting to be sold. Inventory sits as an asset on the balance sheet until it is sold, and it ties up cash while it sits., having never taken delivery of anything. Setu Bazaar carries no cost of goods soldThe direct cost of the items a business sold: what it paid for them, or what it cost to make them. A business that never takes title to what it moves has no line of this kind at all., never having sold goods. Its cost structure is people, engineering, payments handling and the cost of getting each side to show up, none of which scales with the value of the parcels. And its revenue line, the one a reader sees first, is measured in a completely different currency from the number the business talks about when it describes its size.

The second number has a name. Gross merchandise valueEverything bought through a marketplace in a period, added up at the prices the buyers paid. Gross merchandise value measures how busy the place was, and nothing else. is the total value of everything that changed hands through the marketplace in the period, priced at what the buyer paid. Gross merchandise value is a genuinely useful measure of how busy the place is. The figure is not a measure of the business, and the gap between the two can be an order of magnitude wide. Setu Bazaar shows exactly that gap.

Setu Bazaar, described twice, with every component published

Fifty thousand buyers use Setu Bazaar in a year, and each of them transacts Rs 1,00,000/- of goods. Those transactions give Rs 500 crore of goods moving through the marketplace in the year. Setu Bazaar's share of everything that moves through it is 4.00 per cent, so it retains Rs 4,000/- out of each buyer's Rs 1,00,000/-, and Rs 20 crore across all fifty thousand of them. The stock was theirs and the sale was theirs, so the merchants keep the other Rs 480 crore. Every figure can be rebuilt from the two components: the flow, and the share kept.

Read the two bold numbers together. Rs 500 crore and Rs 20 crore describe the same business in the same year, and they differ by a factor of exactly 25.00. Neither is wrong. Neither is the other one dressed up. The two figures answer different questions: how much trade happened, and how much of it did this firm earn. A Rs 500 crore business and a Rs 20 crore business can be the same company, and calling the flow revenue overstates the firm twenty five times over. The measure that connects the two, and the arithmetic for reading it properly, is covered under Take Rate: What a Platform Keeps of What Passes Through. The two numbers exist side by side and are not interchangeable.

How does the value on one side depend on the other?

The dependency between the two sides is the mechanism proper, and it is the reason a platform behaves unlike anything else. Ask what a buyer actually gets from Setu Bazaar. Not a product: Setu Bazaar has no products. The buyer gets access to two thousand merchants. Double the merchants and the buyer's position improves without the buyer doing anything at all. Now ask what a merchant gets. Not a supplier: Setu Bazaar supplies nothing. The merchant gets access to fifty thousand buyers. The value of one side is set by the size of the other side, not by its own size, and that single dependency is what a two-sided market is for.

Notice the shape of that sentence carefully. The claim is a dependency and not a promise. Saying that each side is worth more to the other as the other grows is a statement about direction, not about whether either side will grow, how fast, or whether anyone will pay for the improvement. The economics of markets with two dependent sides were given their formal treatment by Jean-Charles Rochet and Jean Tirole in 2003, and the older idea underneath it, that a thing gets more useful to each holder as more people hold it, was set out by Michael Katz and Carl Shapiro in 1985. Neither result says a platform will grow. Both say what happens to each side's position if it does.

How Platform and Ecosystem Businesses Create Value

A contribution nobody can name leaves the impression that a platform earns for doing nothing. Be concrete about what a firm that makes nothing actually contributes. There are four things, and Setu Bazaar does all four without taking title to a single ream of paper.

The first is matching. A buyer in one town wants a specific thing, and somewhere among two thousand merchants there is one who has it, at a price the buyer would accept. Neither of them could find the other from a standing start. The match itself is the product, and it did not exist before Setu Bazaar made it. The second is the cost of searching. Without the marketplace, fifty thousand buyers would each have to ring round shops, and two thousand merchants would each have to advertise into the dark. One place to look replaces both of those piles of effort at once, and the saving is real even for the trades that would have happened anyway.

The third is terms and trust, and it is the one people undervalue. A buyer is being asked to send money to a stranger in another town. Goodwill is not what makes that possible. The rule releasing the money to the merchant was written down before the trade, applies to everybody, and is enforced by a party with no stake in this particular parcel. The fourth is shared plumbing: payment handling, delivery tracking, a route for disputes. Each is built once and used by every trade on the marketplace. The two thousandth merchant therefore costs Setu Bazaar far less to serve than the first one did. A platform creates value by making a match that would not otherwise have happened and by making it safe to complete, and it is paid out of the value of that match rather than out of a margin on goods.

Why is a platform so hard to start?

Turned round, the dependency becomes the problem that kills most attempts. If a buyer's reason to be here is the merchants, and a merchant's reason to be here is the buyers, then on day one there is no reason for anybody to be anywhere. Neither side will come for an empty other side, and that circle is the cold start problem. It is not a marketing difficulty. The cold start problem is the mechanism running in reverse, and it is the reason a platform's first year looks nothing like its fifth.

Ordinary life shows what this feels like. A new market ground opens on the far side of town on a Thursday. Nobody knows whether anybody will be selling, so nobody walks over on the first Thursday. The sellers have no idea whether anybody will walk, so they do not carry their stock over either. The ground stays empty on a Thursday not because it is a bad ground but because nobody can afford to go first.

There are two ordinary answers, and both of them are ways of paying for somebody to go first. The first is to pay a subsidyMoney one party puts up so that another party faces a lower price than the true cost. Here it means the platform absorbing a cost that a participant would otherwise pay, in order to get that participant to show up. to one side: carry the cost of onboardingEverything it takes to get a new participant set up and actually using a service: signing them up, checking who they are, loading their details, and getting them to their first transaction. merchants, charge them nothing until there are buyers to sell to, and accept that the money goes out before any of it comes back. The second is to start inside a slice narrow enough that a small absolute number is already enough: one town and one kind of goods, where two hundred merchants is not a thin marketplace but a complete one. Both answers buy the same thing: enough liquidityIn a marketplace, the state where somebody who arrives wanting to trade can reliably find a counterparty quickly. A marketplace with participants but no reliable matches is not yet liquid. on one side that the other side has a reason to arrive.

Do all the effects on a platform run the same way?

No, and this is the part that gets left out of almost every short description of a platform. The dependency across the two sides runs positive: more merchants is better for buyers, more buyers is better for merchants. The effect within a side runs the other way. More merchants means more merchants competing for the same buyers, and for a merchant already there that is not an improvement. Cross-side and same-side effects run in different directions, and a reader who knows only the positive one will overstate every platform.

Take the arithmetic rather than the adjective. Hold Setu Bazaar's fifty thousand buyers exactly where they are and double the merchants from two thousand to four thousand. Nothing else changes: the same buyers, the same Rs 1,00,000/- each, the same Rs 500 crore of flow, the same 4.00 per cent kept, the same Rs 20 crore of revenue. Setu Bazaar's own accounts do not move at all. But a buyer now chooses between four thousand merchants instead of two thousand. The same fifty thousand buyers are divided across twice as many shops, so a merchant now reaches 12.50 buyers instead of 25. The flow one merchant handles falls from Rs 25,00,000/- to Rs 12,50,000/- for exactly the same reason.

Read that again with the question who attached to it. The same event, described in the same numbers, is twice the choice for a buyer and half the reach for a merchant. Neither reading is spin. Both are arithmetic. A sentence that says a platform got better has not yet said better for whom, and until it does it has not made a checkable claim.

What does a two-sided structure not establish?

Quite a lot, and the list is worth keeping because a structure is easy to mistake for a result. Setu Bazaar meets all three parts of the test, and that establishes what kind of business it is. The test does not establish that the business makes money, that it will keep its position, or that it captures the value it creates. A structure is not a result, and every one of those three has to be checked separately against figures the structure itself does not supply.

Profitability is the first gap. The costs of getting both sides to show up sit outside the arithmetic of any single trade, so a platform can run a large flow, keep a sensible share of it, and still spend more than it keeps. Whether the money works at the level of one customer is covered under Unit Economics: Profitability at the Level of One Customer, and it is a genuinely separate question from the one worked here. Durability is the second. Nothing stops a merchant from listing on three marketplaces at once, which is multi-homingA participant using several competing services at the same time rather than committing to one. A merchant listing on three marketplaces at once is multi-homing, and it makes each of those marketplaces easier to leave., and nothing stops a buyer and a merchant who met on the marketplace from doing their next trade directly and cutting it out, which is disintermediationTwo parties who were introduced by a middle party dealing with each other directly afterwards, so the middle party stops being paid for the connection it made.. Both are ordinary, both are the reason a position that looks locked can loosen, and neither is visible in the structure.

Capture is the third and the least obvious. Fifty thousand matches happened that would not have happened otherwise, so Setu Bazaar plainly creates value. How much of that value stays with Setu Bazaar rather than passing to the buyers as lower prices or to the merchants as more sales is a completely separate question, answered by where the money in a field actually collects. Capture is covered under The Profit Pool: Where the Money in an Industry Actually Sits. The prior question of whether the field is one where anybody makes money at all is covered under Industry Structure and Sector Behaviour.

How does a practitioner test whether a business really is two-sided?

A business describing itself will always find a way to sound like a marketplace, so the three-part test is easy to nod along to and slightly harder to apply. Here is the single question that does the work, and it is the one an analyst or a lender reaches for first. The question is whether the other side would still have a reason to be here if one side vanished tomorrow. If the answer is yes, the business is not a two-sided market, whatever it calls itself.

Run on Setu Bazaar, the test goes like this. Take away every merchant overnight. Do the fifty thousand buyers still have a reason to open it? No, there is nothing there. Take away every buyer. Do the two thousand merchants still have a reason to list? No, they would be shouting into an empty room. Both directions fail, and that is the mark of the real thing: the two sides are genuinely holding each other up. Run on Anjani Stationers, it comes out otherwise. Take away its paper mill. Do its business buyers still have a reason to buy from Anjani? Yes, entirely, as long as Anjani finds paper somewhere else. The mill is a supplier and it is replaceable. A side of a market is neither of those things.

Each kind of reader then does something different with the answer. There is no inventory to fall back on and nothing to take security over, so a lender looking at a marketplace stops asking about stock and starts asking what happens to the fee income if one side leaves. The ratio between the two numbers is a decision the business made and can change, so an analyst reads them separately and never lets the flow stand in for the revenue. Somebody deciding whether to list their own shop on a marketplace reads the merchant side alone and asks how many buyers they will reach after the marketplace signs up the next thousand shops. And a household choosing where to buy something is, without thinking about it, using the buyer side of exactly the same dependency: they go where the sellers are.

Setu Bazaar and Anjani Stationers Private Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.