For most of business history, the winning move was obvious: control more of the chain. Buy the suppliers, own the factory, run the shops, capture the margin end to end. Then a different kind of company started beating those firms without owning the thing they sold. Airbnb owns no rooms. Uber owns no cars. The app store writes none of the apps. They win by connecting two groups who want each other and taking a cut of the value created when they meet. That is a platform, and it plays by different rules.

The quick version

  • A pipeline business makes something and sells it down a line. A platform business connects producers and consumers and lets them transact, its key asset is the community, not the inventory.
  • Platforms run on network effects: each new user makes the thing more valuable for everyone else, which is why winners tend to take most of the market.
  • The hard part is the start, the chicken-and-egg problem. Neither side shows up for an empty room. You win it by building one tiny, complete network first, not by launching everywhere at once.
  • Platforms aren't always the answer. They're slow and costly to ignite, easy to leave (users multi-home), and they invite regulators. Pick the model the market actually rewards.

The idea in depth

The cleanest framing comes from Marshall Van Alstyne, Geoffrey Parker and Sangeet Paul Choudary in "Pipelines, Platforms, and the New Rules of Strategy" (Harvard Business Review, April 2016). A pipeline creates value in a straight line, design, build, sell, and the firm controls the steps. A platform creates value by orchestrating an interaction between outside producers and outside consumers. The authors' sharp line is that competitive advantage shifts "from controlling resources to orchestrating them," and that the platform's critical asset sits outside the company: the community and the resources its members bring.

That isn't a vocabulary change; it flips three management instincts. Where a pipeline optimises an internal process, a platform curates an external market. Where a pipeline guards its supply chain, a platform's biggest threats can come from a side it doesn't control, including its own producers, who can defect or build a rival. And where a pipeline's strategy is mostly differentiation, a platform's is mostly about getting big enough that the network itself becomes the product. So the move is: before you write a strategy, decide which animal you are. If your advantage comes from what you make, run a pipeline playbook. If it comes from who you connect, almost every default, pricing, the metric you watch, the moat you build, needs rethinking.

flowchart LR
  subgraph Pipeline
    direction LR
    S1("Supplier") --> F("Firm makes it") --> C1("Customer")
  end
  subgraph Platform
    direction LR
    P("Producers") <--> PL(["The platform connects"]) <--> CN("Consumers")
  end
A pipeline pushes value down a line; a platform sits in the middle and lets two sides find each other. Leaders Loop

Why platforms get so big: network effects

The engine underneath all of this is the network effect, a product gets more valuable to each user as more users join. One fax machine is useless; a million is a network. The economics were formalised by Jean-Charles Rochet and Jean Tirole in "Platform Competition in Two-Sided Markets" (Journal of the European Economic Association, 2003), work that contributed to Tirole's 2014 Nobel Prize. Their core insight: a platform serving two interdependent groups must choose a price structure, not just a price level. It often makes sense to subsidise one side heavily (free for riders, free for app users) to attract the side the other side wants. Nightclubs do it instinctively, women in free, men pay, because one side is the draw.

Network effects are also why platform markets tend toward a few big winners rather than many similar players. As Andreessen Horowitz's D'Arcy Coolican and Li Jin lay out in "The Dynamics of Network Effects", the value-per-user curve keeps rising with scale, so the leader pulls away and late entrants face a market that's already good because it's already big. So the move is: stop pricing for short-term margin on both sides and start asking which side is the magnet. Subsidise the magnet, even give it away, and charge the side that will pay to reach it. Then watch engaged, connected users, not raw sign-ups, because an unconnected user generates no network value.

The honest limitation: network effects are weaker and leakier than the hype suggests. Many users multi-home, a driver runs Uber and a rival at once, a diner keeps three delivery apps, which blunts winner-take-all and keeps switching cheap. In The Business of Platforms (HarperCollins, 2019), Michael Cusumano, Annabelle Gawer and David Yoffie show that most platforms fail, and that durable ones usually combine network effects with something stickier (data, switching costs, genuine differentiation). A network effect is an advantage to be earned and defended, not a law that guarantees you the market.

The first problem is the only problem: chicken-and-egg

Every platform launches into an empty room. Consumers won't come without producers; producers won't come without consumers. The two-sided-market economists call this the chicken-and-egg problem, and it's where most platforms die, not in competition, but in failing to start at all. The most useful practical answer comes from Andrew Chen's The Cold Start Problem (2021), drawn from how Uber, Tinder, Slack, Airbnb and others actually got going. His unit of progress is the atomic network: the smallest group that is stable and can grow on its own, for Slack, a single team of three; for a marketplace, one city, one campus, one category.

Chen's second idea is the hard side: in any two-sided network one side is harder to get and to keep (the drivers, the hosts, the sellers, the creators), and they do the work that makes the network valuable. Win them first, with a product that solves a sharp problem for them, and the easy side follows. So the move is: don't launch your platform nationally. Pick the single smallest network that could be complete on its own, throw disproportionate effort at the hard side of it, and get that one network humming before you copy-paste it to the next. A platform that's mediocre everywhere loses to one that's excellent in one place and spreading.

flowchart TD
  A(["Pick one tiny market
(a city, a campus, a niche)"]) --> B("Win the HARD side first
sellers, hosts, drivers")
  B --> C("Easy side follows
buyers, guests, riders")
  C --> D{"Self-sustaining
atomic network?"}
  D -->|"Not yet"| B
  D -->|"Yes"| E("Copy it to the
next market")
Igniting a platform: build one complete network, win its hard side, then repeat, rather than launching everywhere thin. Leaders Loop

A worked example

Picture "RoundTrip," an imaginary app for booking mobile mechanics, drivers who'll come to your driveway. (Figures below are illustrative, to show the reasoning, not real data.)

The founder's instinct is to launch in five cities and spend the marketing budget on car owners, because owners are easy to reach with ads. Six weeks in, owners open the app, see two mechanics, neither available this week, and never return. Money spent, nothing ignited, the empty-room problem in miniature.

The platform reframing changes the plan. One network, not five: RoundTrip pulls back to a single suburb of 40,000 households. Hard side first: the constraint isn't owners, there are plenty, it's vetted mechanics willing to travel, so the founder personally recruits twelve good ones and guarantees each a minimum number of paid jobs for the first month. Price structure, not price level: RoundTrip stays free for owners (the magnet the mechanics want) and takes a 15% cut from mechanics (the side that earns from the connection). Now an owner opens the app and sees eight mechanics free this week, a network worth returning to. Then, and only then, the founder clones the model into the next suburb. The advantage was never the app, which is ordinary software. It was being the first place mechanics and owners reliably found each other.

Frequently asked questions

What's the difference between a platform and an ecosystem?

A platform is the structure that connects two or more sides and lets them transact. An ecosystem is the wider web of partners, complements and third parties that grow up around it, the app developers, accessory makers and integrators who add value the platform owner never built. A console is a platform; the studios, peripheral makers and streamers around it are its ecosystem. Strong platforms deliberately cultivate that ecosystem, because complements they don't have to build make the platform more valuable.

Does every business need a platform strategy?

No, and forcing one is a common, expensive mistake. Platforms only work where two groups genuinely need a matchmaker and where the connection creates real value. If your edge is craftsmanship, brand or a proprietary process, a pipeline can be the stronger model, and many platform attempts fail precisely because they bolted a marketplace onto a problem that didn't have two sides. The Business of Platforms is blunt that most platforms lose money and shut down; the question is whether the market is shaped like a platform, not whether platforms are fashionable.

How do you beat a platform that already has the network effect?

Rarely head-on. The usual openings are: multi-homing (users happily run several apps, so you ride alongside the incumbent rather than replacing it); a distinct sub-segment the giant serves poorly, where you build your own atomic network; or a different price structure or trust model that peels off one side. You don't out-scale a network leader on day one, you find a corner of the market where their size isn't the thing that matters.

Why give one side the product for free?

Because in a two-sided market the sides value each other, not just the platform. If one side is the reason the other shows up, subsidising or free-pricing the magnet side and charging the side that profits from reaching it can grow the whole network faster than charging both. The art is choosing the right side to subsidise, get it backwards and you pay to attract people nobody else wants.

What's the single most common reason platforms fail?

They never solve the cold start. Founders launch broad and thin, money runs out before either side reaches the density that makes the product useful, and the network never reaches the point where it grows itself. Winning one small, complete network first is unglamorous and slow, and it's the difference between a platform that ignites and one that quietly empties out.

Related in the Toolkit

Where to go next