What Are Two-Sided Markets?

A two-sided market is a market where a platform serves two distinct groups, and each group's benefit from joining depends on how many members of the other group join. Cardholders want merchants who accept the card, and merchants want cardholders who carry it. Readers want a newspaper with good journalism, and advertisers want a newspaper with readers.

That interdependence changes almost everything about how such a business has to be run. Prices can't be set side by side the way a bakery sets the price of bread. Starting up is harder, because neither group moves first. And competition can end differently, with one winner or several, depending on a few structural conditions.

This article is the economics concept: what the theory says, who developed it, and which parts hold up as practical guidance. It isn't a guide to building a marketplace or a platform. For the business-model versions, read the marketplace business model and platform business model articles. This one explains the theory underneath both.

Where the idea comes from

The formal literature took shape in the early 2000s. Jean-Charles Rochet and Jean Tirole's 2003 paper, "Platform Competition in Two-Sided Markets" in the Journal of the European Economic Association, opens with the observation that many if not most markets with network externalities are two-sided. Its abstract says platforms in industries such as software, portals and media, and payment systems must "get both sides of the market on board," and that they therefore devote much attention to how they court each side while making money overall.

Several related papers appeared close together:

A practitioner-friendly version arrived in Harvard Business Review. Eisenmann, Parker and Van Alstyne's "Strategies for Two-Sided Markets" (October 2006) says two-sided networks, or platforms, often enjoy significantly greater returns at scale than traditional value chains. It gives the credit card as its lead example, and adds newspapers (which join subscribers and advertisers), HMOs and computer operating systems.

For context, the Royal Swedish Academy of Sciences awarded the 2014 Prize in Economic Sciences to Jean Tirole "for his analysis of market power and regulation". The motivation doesn't name two-sided markets, so it isn't an endorsement of that specific paper. It does show the theory sits inside mainstream industrial economics.

The definition that matters: price structure

Plenty of businesses have two customer groups. A shop has suppliers and shoppers, and nobody calls it a two-sided market. What's special?

Rochet and Tirole's 2006 progress report gives the sharpest answer. Its abstract says they identify two-sided markets with markets in which the structure, and not only the level of prices charged by platforms, matters. It adds that the failure of the Coase theorem is necessary but not sufficient for two-sidedness.

In plain terms: in an ordinary market, what matters is the total price. In a two-sided market, how that price is divided between the two groups changes how many transactions happen. Moving a fee from one side to the other can change volume even when the total stays the same. The Coase-theorem point is the economist's way of saying the two sides can't bargain their way to the best split on their own, which is one reason a platform has to set the split for them.

That gives a working test for any business:

1. Are there two distinct groups?
2. Does one group's value from the platform depend on how many of the other group take part?
3. Does the way you split fees between the groups change how many transactions happen?

If all three are yes, the pricing and growth advice for ordinary businesses will often mislead you.

Same-side and cross-side network effects

A network effect means a product gets more valuable as more people use it. Two kinds appear in two-sided markets, and the network effects article covers both in more depth.

• Same-side effects run within one group. More users of a messaging service make it better for the other users of that service. They can also be negative: more competing sellers on a platform means each seller gets less attention.
• Cross-side effects run between groups. More merchants accepting a card make it more valuable to cardholders, and more cardholders make it more valuable to merchants.

The US Supreme Court described the cross-side version in its 2018 opinion in Ohio v. American Express Co.: "the value of the two-sided platform to one group of participants depends on how many members of a different group participate." For credit cards specifically, the opinion says a card is more valuable to cardholders when more merchants accept it, and more valuable to merchants when more cardholders use it. It also notes that a transaction platform can't sell to one side without simultaneously selling to the other.

Cross-side effects can turn negative too. A newspaper's readers may like some advertising and dislike a lot of it.

Subsidy side and money side

If prices matter structurally, which side should pay? Rochet and Tirole's 2003 abstract says their model identifies the determinants of price allocation. Armstrong's 2006 abstract lists the factors in plain language: the magnitude of the cross-group externalities, whether fees are levied on a lump-sum or per-transaction basis, and whether agents join one platform or several.

From those results, practitioners describe a subsidy side and a money side. The subsidy side is the group the platform charges little or nothing (and sometimes pays), because that group's participation is what makes the platform valuable to the other. The money side is the group that pays more because it values access to the first.

Parker and Van Alstyne's model gives a formal reason a firm might do this. Their abstract states that even without competition, a firm can rationally invest in a product it intends to give away into perpetuity, that either the content-provider side or the consumer side can be the candidate for a free good, and that the model offers insights to regulators applying antitrust law to network markets.

A few illustrations, described qualitatively:

These are patterns the literature discusses, not fixed rules, and platforms change their splits over time. Check a company's own filings before stating who pays what.

The chicken-and-egg problem

Caillaud and Jullien's 2003 paper gave the startup problem its best-known name. Because each side wants the other, a new platform faces a coordination failure: nobody joins an empty network.

The standard responses are tactics rather than guarantees, and the theory explains why each works:

• Subsidize the side that the other side values most. This follows directly from the pricing logic above.
• Start narrow. A single city, category or community can reach a useful density sooner than a nation can.
• Build one side's value without the other. A tool that's useful to a single participant lets you collect that side first.
• Seed supply manually. Early listings or partners make the first visit worthwhile.
• Borrow an existing audience. Launch inside a community where one side already gathers.

For cold starts in practice, including liquidity and leakage, see the marketplace business model article.

Multi-homing and competitive bottlenecks

Multi-homing means a participant uses more than one competing platform at the same time. Single-homing means joining only one. The distinction matters because it decides where a platform's power sits.

Armstrong's 2006 paper includes a model of "competitive bottlenecks," where one group joins all platforms. Per his abstract, equilibrium prices in these models depend on the size of the cross-group externalities, on whether fees are lump-sum or per-transaction, and on whether agents join one platform or several. So whether each side single-homes or multi-homes is a direct input to who pays, not a footnote.

Caillaud and Jullien add a related result about non-exclusivity. Their abstract says intermediaries have incentives to propose non-exclusive services, as this moderates competition and allows them to exert market power. In other words, letting participants use rivals as well isn't always a sign of weakness. It can soften competition.

When do two-sided markets become winner-take-all?

Network effects pull toward one big platform. The same literature shows the pull can be resisted. Whether a market ends up concentrated depends on conditions that work together:

This is a pattern across the theory, not a law, and the papers cited above model specific cases. Be wary of any claim that two-sided markets "always" end in monopoly. The practical use of the table is diagnostic: before assuming your market will tip, check the first two rows. If both sides can easily use several platforms, a strong cross-side effect alone won't hand you the whole market. For how this relates to defensibility, see the economic moat and first-mover advantage articles, and for how network effects feed growth loops, viral network effects.

Why courts and regulators care

Two-sidedness has legal consequences, because ordinary tests for harm can mislead when they look at only one side.

In Ohio v. American Express (2018), the Supreme Court held that courts must include both sides of the platform, merchants and cardholders, when defining the credit-card market, because transaction platforms are better understood as supplying only one product, transactions. The Court affirmed that American Express's antisteering provisions did not violate federal antitrust law, finding the plaintiffs hadn't shown anticompetitive effects when the market was analyzed as a whole instead of looking at merchant fees alone. It addresses a transaction platform, so it doesn't automatically apply to every platform business.

In the EU, the Digital Markets Act takes a different tack. The European Commission describes it as the EU's law to make the markets in the digital sector fairer and more contestable, applying to large "gatekeeper" platforms providing core platform services such as online search engines, app stores and messenger services. The page doesn't discuss two-sided market theory directly, so treat the DMA as the regulatory response to platform power rather than as an application of the theory.

The takeaway: how you define a market, one side or both, can decide who wins an argument about prices and harm.

Using the theory without overreaching

For a founder, the theory is most useful as a set of questions, not formulas:

1. Which side is harder to get, and which side does it value? That's the likely subsidy side.
2. Do both sides care about the other's size, or only one? One-way effects need a different strategy from two-way ones.
3. Can participants multi-home cheaply? If yes, defensibility must come from something other than size.
4. Do the numbers still work? Subsidies cost money, so check unit economics per participant on each side.

The economics tells you the shape of the problem, not whether a given platform will solve it.

Key Facts

• A two-sided market links two groups where one group's benefit from joining depends on how many of the other join, as in the Supreme Court's description in Ohio v. American Express.
• Rochet and Tirole's 2006 progress report identifies two-sided markets with those where the structure, and not only the level, of prices matters.
• Rochet and Tirole's 2003 paper says platforms must "get both sides of the market on board".
• Armstrong's 2006 paper models a "competitive bottleneck," where one group joins all platforms and prices depend on cross-group externalities, fee type and homing.
• Caillaud and Jullien's 2003 paper, Chicken & Egg, models indirect network externalities and shows intermediaries can profit from non-exclusive services.
• In 2014 Jean Tirole received the economics prize "for his analysis of market power and regulation".
• Concentration depends on network effects, multi-homing, switching costs and differentiation together, so winner-take-all is a tendency, not a rule.

Related Reading

• Startup Business Models: The Main Types Explained
• Marketplace Business Model
• Platform Business Model
• Network Effects
• Viral Network Effects
• Economic Moat
• First-Mover Advantage
• Unit Economics