Uber doesn't own cars. Airbnb doesn't own rooms. Visa doesn't make anything you can hold. What each one actually sells is a match, a rider to a driver, a guest to a host, a shopper to a shop, and the price it charges has a logic that looks bizarre until you understand the one rule underneath it: on a platform, the two sides need each other, and you are pricing the relationship, not the product.

The quick version

  • A two-sided (or multi-sided) platform creates value by connecting two distinct groups who each want the other, buyers and sellers, riders and drivers, users and developers.
  • The engine is the cross-side network effect: more of one side makes the platform more valuable to the other. That feedback loop, once it spins, is the moat.
  • Because the sides depend on each other, you often subsidise one side to attract the other, the price structure (who pays what) can matter more than the total price.
  • The catch is chicken-and-egg: neither side shows up until the other is already there. Solving cold-start is the whole game in year one.

The idea in depth

Economists have a precise test for what makes a market "two-sided," and it's sharper than the everyday use of the word. Jean-Charles Rochet and Jean Tirole, Tirole won the 2014 Nobel in economics partly for this work, argued in their foundational papers that a market is two-sided when the volume of transactions depends not just on the total price the platform charges, but on how that price is split between the two sides. Shift a dollar of the fee from sellers to buyers and the number of deals changes, even though the platform's total take is identical (Rochet & Tirole, "Two-Sided Markets: A Progress Report," 2006). In a normal market, only the total price moves behaviour. On a platform, the structure of the price does too.

That single property is why platforms behave so differently from ordinary businesses, and it's worth holding onto, because every move below falls out of it.

flowchart LR
  A(["Side A, e.g. riders, guests, buyers"]) -- "demand" --> P(["Platform  (the match-maker)"])
  B(["Side B, e.g. drivers, hosts, sellers"]) -- "supply" --> P
  P -- "more of B makes A happier" --> A
  P -- "more of A makes B happier" --> B
A platform sits between two groups and gets more valuable to each as the other grows. Leaders Loop

Network effects: the loop that becomes the moat

The reason the price split matters is network effects, the property that a platform gets more valuable as more people use it. Geoffrey Parker, Marshall Van Alstyne and Sangeet Paul Choudary, in Platform Revolution (2016), call demand-side economies of scale the defining force of platform businesses: value comes from who else is on the network, not from making each unit cheaper to produce (Parker, Van Alstyne & Choudary, Platform Revolution, 2016).

It helps to split the effect in two. Cross-side effects run between the groups: more drivers means shorter waits for riders; more riders means more fares for drivers. Same-side effects run within a group, and can be positive (more developers on a console means more shared knowledge) or negative (more drivers competing for the same riders thins everyone's income). Getting these signs right per side is the core design question.

So the move is: for your platform, draw a two-by-two of every side against every other side and label each cell with the sign of the effect. Where it's strongly positive, you have a growth flywheel to lean on. Where a same-side effect is negative, too many sellers crowding each other out, that's where you'll need rules, ranking, or supply caps so growth doesn't quietly degrade the product. This is also where the moat lives: a competitor can copy your features in a weekend, but they can't copy the fact that everyone is already on your side of the street. As Parker and his co-authors put it, that's how networked markets flip the old logic of scale.

A rival can copy your features. They can't copy the fact that the other side is already here.

An honest limitation. Network effects are not destiny. They can be local rather than global, Uber's liquidity in one city does nothing for a rider in another, which is why ride-hailing was fought city by city, not won once. They can reverse: when quality falls, the same loop that filled the platform can empty it (the "anti-network effect"). And they don't stop multi-homing, drivers run Uber and a rival app on the same windscreen mount, blunting any single platform's lock-in. Treat the flywheel as something you have to keep spinning, not a switch you flip once.

Pricing: why one side rides for free

Here's where the Rochet–Tirole insight earns its keep. Because the sides value each other, you can charge one heavily and the other barely, or even pay the second side to show up. Thomas Eisenmann, Geoffrey Parker and Marshall Van Alstyne, in their Harvard Business Review article "Strategies for Two-Sided Markets" (2006), call this picking the subsidy side and the money side: you subsidise the group that is more price-sensitive and more attractive to the other, then earn your margin from the group that will pay to reach them (Eisenmann, Parker & Van Alstyne, HBR, Oct 2006).

It's everywhere once you notice it. Adobe gives the Reader away and charges for Acrobat. Nightclubs let women in free to draw a paying crowd. Search engines and social networks are free to you because you are the side advertisers pay to reach. The free side isn't generosity; it's the bait that makes the other side worth charging.

So the move is: before you set a price, decide which side is the subsidy side. Ask two questions about each group, how price-sensitive are they? and how much does the other side want them? Subsidise the sensitive-but-attractive side; monetise the side that needs access and can pay for it. Then check whether being in the payment flow is even possible: the venture investor Bill Gurley notes it is far easier to capture reasonable economics when the transaction runs through you rather than around you (Gurley, "All Markets Are Not Created Equal," 2012). This is the bridge into value-based pricing and the deeper unit economics of each side.

An honest limitation. Subsidies bleed cash, and a subsidised side can be the wrong side. If you give away the part that drives your real costs, growth makes the hole deeper, the more it's used, the more you lose. Several well-funded marketplaces have scaled themselves into trouble by subsidising demand they could never profitably serve. The subsidy is an investment in igniting the loop, not a permanent business model; you need a credible path to the money side carrying the whole thing.

That the two sides are economically inseparable isn't just a strategy point, it's now law. In Ohio v. American Express (2018), the US Supreme Court held that a credit-card network is a single, two-sided "transaction" market and that courts must weigh effects on both cardholders and merchants together, not one in isolation (Ohio v. American Express Co., 585 U.S. 529, 2018). When the country's highest court reorganises antitrust around the idea, you can take the model seriously.

A worked example

Picture a founder launching "Brief," a marketplace matching small businesses with freelance designers. (Figures below are illustrative, for teaching only.) The two sides: businesses wanting design work, and designers wanting paid gigs. The cross-side effect is obvious, more designers means more choice and faster turnaround for businesses; more businesses means more income for designers.

The chicken-and-egg problem hits on day one. A business opens the app, sees three designers, and leaves. A designer signs up, sees no briefs, and never returns. Neither side will wait for the other. So the founder does what Platform Revolution calls seeding the harder side: she personally recruits 50 strong designers, paying a small retainer so portfolios are live before any business arrives, the marketplace revenue model deliberately running at a loss to bootstrap supply.

flowchart TD
  S(["Seed the hard side: 50 designers paid to list"]) --> L(["Businesses see real choice → post briefs"])
  L --> M(["Designers earn → invite peers, stay active"])
  M --> N(["More supply → faster, cheaper matches"])
  N --> L
  N --> X(["Flywheel spinning: each side pulls the other"])
Cracking cold-start by subsidising the side that's harder to attract, then letting the loop self-feed. Leaders Loop

On pricing, she names the subsidy side (designers, fee-sensitive, and the scarce side businesses are hungry for) and the money side (businesses, who'll pay to get work done). Designers list free; businesses pay a 12% fee on completed projects, captured because payment flows through Brief. As liquidity builds, the negative same-side effect appears, too many designers chasing each brief, so she adds curated ranking and a "verified" tier rather than letting volume erode quality. Each move falls straight out of the model.

Frequently asked questions

What's the difference between a platform and a normal marketplace or pipeline business?

A "pipeline" business creates a product and sells it down a line to a customer (a factory, a shop, a consultancy). A platform doesn't create the core value itself, it lets two outside groups create value for each other and takes a cut of the connection. The economic tell is the Rochet–Tirole test: if shifting the fee from one side to the other changes how much gets transacted, you're running a platform, not a pipeline.

Why do platforms give one side the product for free?

Because the free side is what makes the paying side worth charging. Advertisers pay to reach users; merchants pay to reach cardholders; businesses pay to reach designers. You subsidise the group that is price-sensitive and attractive to the other, and you earn from the group that will pay for access (Eisenmann, Parker & Van Alstyne, 2006). "Free" is a customer-acquisition strategy aimed at the other side's wallet.

How do you solve the chicken-and-egg problem?

Common, evidence-backed tactics: seed the harder-to-attract side first (often supply), even at a loss; go narrow to reach liquidity in one niche or one city before expanding; provide standalone value so early users get something even before the other side arrives; and, where you can, do things that don't scale by hand-recruiting the first cohort. The goal is local liquidity, not national coverage, a thin platform everywhere is worse than a dense one somewhere.

Are network effects really a durable moat?

Sometimes. They're strongest when effects are global rather than local, when switching is costly, and when users can't easily multi-home. They're weak or reversible when liquidity is local (you re-fight every city), when rivals sit on the same users' phones, or when quality slips and the loop runs backwards. Treat the moat as conditional, not guaranteed.

Can a platform have more than two sides?

Yes, these are multi-sided platforms. A games console serves players, developers and sometimes accessory makers; a payment card serves cardholders, merchants and issuing banks. The same logic scales: map every side against every other, find the cross-side effects, and decide who you subsidise and who you charge.

Related in the Toolkit

Where to go next