Pattern · Strategy & competition
The Multi-Sided Platform
Also known as Two-sided market, Platform business, Matchmaker. Anchored to Jean-Charles Rochet and Jean Tirole, 2003.
A business that makes money by bringing two groups together who each need the other, where the decisive question is not what to charge but which side to charge, because the side you subsidise is the side that makes the whole thing work.
Plate · The shape
How it works
Rochet and Tirole's contribution was precise and is often lost in the retelling. A market is two-sided when the platform can change the total volume of business by shifting the price between the two sides while leaving the total price unchanged. That is the test. If moving the charge from buyers to sellers, or the reverse, changes how much trade happens, you are in a two-sided market and the price structure matters as much as the price level.
The reason it matters is that each side values the platform according to how many of the other side are on it. Nightclubs let women in free. Payment networks charge merchants and pay cardholders. Newspapers charge readers a fraction of production cost and bill advertisers for the audience. In each case one side is subsidised because its presence is what the other side is buying.
The hard part is not the theory. It is the beginning. A platform with nobody on either side is worth nothing to either side, and the founder has to buy, fake or force the first side into existence before the second will arrive. Almost every platform that failed, failed there, and almost every platform that succeeded has an unglamorous story about how it seeded the first side by hand.
The parts of the model
The platform
The rules, the matching mechanism and the trust apparatus that make an exchange between the two sides safe and cheap enough to be worth doing. It is not the software. The software is how the rules are enforced.
Signals
The platform sets terms it can enforce against both sides · Disputes are resolved by the platform rather than between the parties · Removing the platform would raise the cost of the same transaction, not merely the convenience
The subsidised side
The group whose presence the other side is actually paying for. Charged little or nothing, often courted expensively, and usually the side that is more price-sensitive or harder to attract.
Signals
This side would leave for a competitor over a small fee · Its numbers are quoted in every conversation with the other side · The business spends money acquiring it and books that spend as marketing rather than as cost of revenue
The money side
The group that pays, because access to the subsidised side is worth more to them than the fee. Usually less price-sensitive, often fewer in number, and frequently the side that arrived second.
Signals
Willingness to pay tracks the size and quality of the other side, not the platform's features · This side asks for exclusivity, better placement or data rather than a lower price · Losing a member of this side hurts revenue immediately and volume not at all
The network effect
The mechanism by which each side becomes more valuable to the other as it grows. Cross-side effects are usually positive; same-side effects are often negative, since sellers do not want more sellers.
Signals
Growth on one side measurably raises retention on the other · Sellers complain about crowding while buyers do not · Value per user rises with scale rather than merely total value
How you know you are in it
You do not fill a pattern in. You recognise yourself in it, or you do not. Read these as a list about your own business rather than as a definition.
- You have two distinct customer groups and each one asks you how many of the other you have before asking about price.
- One group is billed and the other is courted, and if you tried to charge the courted group even a token amount you are fairly sure most of them would leave.
- Your growth conversations are about chickens and eggs, and you have used that phrase in a board meeting.
- Cost of acquisition on one side is properly a cost of revenue on the other, and your accounts do not show it that way.
- Competitors are copying your features and it is making no difference, because what they cannot copy is who is already standing in your room.
- The most dangerous thing anyone could do to you is persuade both sides to meet somewhere else, and you know exactly who would try.
The numbers that decide it
- The price structure, not the price level. Model total take across both sides, then model what happens to volume if you move a pound of it from one side to the other. If volume moves, the structure is a live strategic lever and should be reviewed as often as the price is.
- Cross-side elasticity: how much does adding a hundred sellers change buyer retention, and the reverse. Most platforms guess this. The ones that measure it know which side to spend on.
- Cost of acquisition on the subsidised side, treated honestly as a cost of serving the money side rather than as marketing. Reclassifying it usually changes the apparent margin substantially.
- Multi-homing rates: what proportion of each side also uses a competitor. High multi-homing on both sides means there is no lock-in and the take rate will be competed down, whatever the network effect story says.
- Liquidity by segment rather than in aggregate. A marketplace can look healthy overall and be dead in every individual category, which is the failure mode most totals hide.
When this shape works
- The two sides genuinely cannot find each other cheaply on their own, so the platform is removing a real search or trust cost rather than inserting itself into an existing relationship.
- One side is clearly more price-sensitive than the other, which makes the subsidy decision obvious rather than a guess.
- Transactions are frequent enough that habit forms. One-off, high-value matches tend to leak the moment the two parties have each other's number.
- The platform can hold something the parties cannot take with them: reputation, payment protection, dispute resolution, or data that only exists because the platform sees both sides.
And when it doesn't
- You are inserting yourself between two parties who already deal with each other directly and are happy about it, which is intermediation rather than matchmaking and is priced accordingly.
- Both sides multi-home cheaply and switching costs are near zero, in which case the network effect is real and worth nothing.
- The business has no way to fund the subsidised side through the years before the money side is large enough to pay for it. This is the specific reason platforms are a poor fit for an owner-managed business without outside capital.
- The valuable transaction is rare and high-value, since both parties have every incentive to complete the second one off-platform and the platform's only defence is a contract nobody enjoys enforcing.
How this pattern dies
The cold start
The commonest death, and it happens quietly. Neither side arrives because the other is not there. The founder spends eighteen months building features for a platform nobody is standing on. The platforms that survived this almost all cheated in the same way: they hand-recruited one side at unsustainable cost, or launched in a single narrow segment small enough to fill.
Disintermediation
The two sides meet on the platform and transact off it. This is not theft by unusual users; it is the rational behaviour of every repeat pairing, and any platform whose value is only introduction will bleed its best relationships first, since the best ones are the ones worth taking outside.
Subsidising the wrong side
The subsidy goes to whichever side the founder finds easier to talk to, which is usually the side he came from. The tell is a platform with plenty of sellers, no buyers and a founder who used to be a seller.
Take-rate creep
Growth slows, the take rate rises to compensate, and the money side begins to model whether it could do this itself. Every platform that has been squeezed by its own supply side got there one percentage point at a time, and the point of no return is invisible until it has passed.
Winner-take-all as an article of faith
The strategy assumes the market tips to one platform and burns capital to be that platform. Whether markets tip depends on multi-homing, differentiation and congestion, and in many markets they simply do not. The assumption is expensive to hold and almost never tested before the money is spent.
In the wild
Rightmove
UK, and the clearest domestic case. Home buyers browse free; estate agents pay substantial monthly fees for listings because the buyers are there. The agents have twice attempted collective alternatives and the audience did not move, which is what a genuine network effect looks like from the inside.
The village auction house
Owner-managed, UK, and centuries older than the theory. Buyers attend free, sellers pay commission, and the auctioneer's whole business is that both turn up on the same morning. Everything Rochet and Tirole formalised was already being run out of a shed in a market town.
Payment card networks
The case that produced the theory. Cardholders are paid to participate through rewards while merchants carry the interchange fee, and the resulting regulatory arguments in the EU and UK are, in substance, disputes about price structure rather than price level.
The trade association with a supplier directory
Owner-managed and unglamorous. Members join for the standards and the networking; suppliers pay for access to a qualified audience. Small, durable, and a reminder that a platform does not have to be software or venture-funded to be a platform.
The owner-managed version
I would be careful with this one. Platforms are the most admired pattern in business writing and the worst fit for most owner-managed firms, because the model requires you to fund one side for years before the other side pays, and that funding has to come from somewhere. Venture capital exists largely to solve exactly this problem. If you do not have it, the cold start will eat your working capital and you will not know it is happening until it has.
That said, there is a version of this that works at small scale and I have seen it done well. Pick a segment narrow enough that you can personally fill one side of it. Not an industry. A town, a trade, a niche where you could name most of the participants. Fill the harder side by hand, on the phone, one at a time, because there is no clever substitute for that and every successful platform did it. Then charge the side that is buying access, and charge them properly, because a platform that undercharges the money side has no way of ever funding the other one.
The question I would put to any owner considering it: what do you hold that they cannot take with them when they meet? If the honest answer is nothing, you are not building a platform. You are building an introductions business, which is a perfectly good business, but it is priced per introduction and it does not compound.
One more thing worth saying, because it is a cheap win. A good many owner-managed businesses are already running a small platform without noticing: a supplier list they curate, an events programme where two groups meet, a referral flow they operate for free. That is a subsidised side already built. The question is whether the other side would pay for organised access to it.
Changing out of this shape
Getting into this pattern deliberately means answering three questions in order and refusing to move on until each is settled. Which side is scarce, since that is the side to subsidise and it is rarely the side you find easiest to sell to. What holds the transaction on the platform once the two parties know each other, since introduction alone never holds. And who funds the subsidy until the money side is large enough, since the answer is either outside capital, an existing profitable business, or a segment small enough to seed by hand. Getting out is harder: a platform that has been subsidising the wrong side for years cannot simply flip the charge without losing the volume the money side was buying, so the usual route is to introduce a paid tier on the subsidised side that sells something other than access, such as tools, placement or data.
Where to go from here
Frameworks that work inside this pattern
- Business Model Canvas
A one-page visual template that describes how an organisation creates, delivers and captures value across nine building blocks, from customer segments through to cost structure, so an entire business model can be seen, questioned and redesigned in a single view.
- Porter's Five Forces
A structural analysis of the five competitive forces that determine an industry's long-run profitability, and therefore where and how a business can defend or improve its position.
- Unit Economics
Tests whether a business makes money on each unit sold and each customer acquired, tracing revenue through variable costs to contribution and comparing the cost of acquiring a customer with the lifetime value they return. If the unit loses money, scale multiplies the loss.
- Jobs to be Done
A lens for understanding demand: customers hire products to make progress in a specific circumstance, and whether they switch is governed by four opposing forces. Study the job and the forces around it rather than the customer's attributes or the product's features.
- Crossing the Chasm
Moore's account of why technology products stall between visionary early adopters and pragmatist mainstream buyers, and the beachhead strategy for getting across: pick one narrow segment, build the whole product it needs, and win the references pragmatists actually trust.
- 7 Powers
Helmer's checklist of the seven conditions that create persistent differential returns. Each power pairs a benefit to the holder with a barrier that stops competitors arbitraging it away, and each becomes available only at a particular stage of a business's life.
Playbooks that work the problem
- Entering a New Market
Four frameworks in the order I would actually use them: read the macro weather, judge whether the industry can pay you, name the risk you are really taking, then make a choice whose reasoning you can still defend a year later.
- Pricing with Confidence
A sequence for putting a number on your work without flinching: find the job customers actually hire you for, price the value rather than the hours, check the volume the price must sustain, then imagine the new price list failed and find out why before it does.
Neighbouring patterns
- The Marketplace Take Rate
Taking a percentage of transactions between two parties you never own the goods for, which produces beautiful margins and no inventory, and leaves the whole business resting on whether the two parties keep coming back through you rather than going round you.
- Free and Freemium
Giving something substantial away at no charge so that somebody else pays for it: an advertiser, a subset of users who upgrade, or the same user later, which works when the free thing costs almost nothing to reproduce and fails quietly when it does not.
- Unbundling
Most companies are three businesses wearing one badge: finding and keeping customers, inventing products, and running infrastructure at scale. Each wants a different cost base and a different pace, so holding all three under one roof means at least two are being run badly.
What the critics say
Winner-take-all is the assumption most often smuggled in with this pattern and it is not supported. Rysman's survey of the empirical literature concludes that whether a two-sided market tips to a single platform depends on multi-homing, differentiation and congestion, and that many such markets sustain several competing platforms indefinitely. Strategy built on inevitable tipping is a bet, not an inference.
Rysman, M. (2009) 'The Economics of Two-Sided Markets', Journal of Economic Perspectives, 23(3), pp. 125-143.
The term has been stretched until it means almost nothing. Hagiu and Wright set out the conditions that actually distinguish a multi-sided platform from a reseller or a vertically integrated firm, chiefly that both sides retain control over key transaction decisions, and note how many self-described platforms fail that test. A business calling itself a platform is not evidence that it is one.
Hagiu, A. and Wright, J. (2015) 'Multi-Sided Platforms', International Journal of Industrial Organization, 43, pp. 162-174.
The theory is largely silent on the transition. Rochet and Tirole model competition between platforms that already exist, which leaves the hardest practical problem, getting the first side onto an empty platform, outside the formal framework entirely. Practitioners consequently take their launch advice from case anecdote rather than from the literature that gave the pattern its name.
Evans, D. S. and Schmalensee, R. (2016) Matchmakers: The New Economics of Multisided Platforms. Boston, MA: Harvard Business Review Press.
Sources and further reading
- Rochet, J.-C. and Tirole, J. (2003) 'Platform Competition in Two-Sided Markets', Journal of the European Economic Association, 1(4), pp. 990-1029.
- Rochet, J.-C. and Tirole, J. (2006) 'Two-Sided Markets: A Progress Report', RAND Journal of Economics, 37(3), pp. 645-667.
- Eisenmann, T., Parker, G. and Van Alstyne, M. W. (2006) 'Strategies for Two-Sided Markets', Harvard Business Review, 84(10), pp. 92-101.
- Parker, G. G., Van Alstyne, M. W. and Choudary, S. P. (2016) Platform Revolution. New York: W. W. Norton.