A network effect exists when a product is worth more to each of its users the more other people use it. The telephone is the standard example: one handset is useless, and each new subscriber makes every existing line more valuable. A platform is a business or protocol whose product is the connection itself. It brings together groups who want to reach each other, such as card holders and merchants, riders and drivers, or phone owners and app developers.
Both share one difficulty. A network is worth joining once enough people are on it, so adoption feeds itself after it starts and stalls before it does, and the firm that gets past the start can end up holding a market its users are reluctant to leave. Economists have studied that problem since the standards wars of the 1990s. Since 2017, blockchain projects have proposed a new answer to the start: give a network's first users a stake in it.
Carl Shapiro and Hal Varian wrote Information Rules (Harvard Business School Press, 1999) as a strategy guide for managers of information businesses, and organised it around two forces. The first is positive feedback: as a network's installed base grows, more people find it worth adopting, and in their words "positive feedback makes the strong get stronger and the weak get weaker, leading to extreme outcomes". Economists call such a market tippy, because it can tip to one standard and orphan the others, as the VHS video cassette format did to Sony's Beta. The second force is lock-in. Once customers have invested in one system, moving to another costs them retraining, converted files and replacement equipment. Shapiro and Varian's rule of thumb is that the profit a supplier can expect from a customer equals those total switching costs, plus whatever advantage its product or its costs give it over rivals.
Luís Cabral's Dynamic Price Competition with Network Effects (The Review of Economic Studies, 2011) puts the tipping story into a formal model and qualifies it. Two firms own rival networks. Consumers arrive, join one of them, and eventually leave, and each firm sets an entry price for newcomers, which may be below cost. Two forces act on that price: a larger network is more attractive, so it can charge more, but a larger firm also has more to gain from growing further, so it has reason to price lower.
The result is a U-shaped pricing function in which very small and very large networks charge the most. Market shares show increasing dominance around an even split, where the larger network tends to pull ahead, and reversion to the mean near monopoly, where it tends to shrink. With strong network effects the market spends most of its time with one network holding between 50 and 100 per cent of it, rather than all of it. Cabral also measures network effects as a barrier to entry: the fall in an entrant's value that their presence causes.
David Evans and Richard Schmalensee, economists who were among the first to analyse multisided platforms, wrote Matchmakers: The New Economics of Multisided Platforms (Harvard Business Review Press, 2016) for the people who build them. A telephone network has a direct network effect: subscribers value other subscribers. A platform runs on indirect network effects, where one group values the platform for the number of members of a different group it attracts: shoppers care how many merchants accept a card, and merchants care how many shoppers carry it. Evans and Schmalensee argue that the first-mover and winner-take-all reasoning developed for technology standards does not carry over to such businesses. What a matchmaker needs first is critical mass: enough participants on every side that each finds the others worth joining.
Two consequences follow. A platform prices to balance its sides rather than to cover its costs, so it often charges one side little or nothing, the subsidy side, and recovers the cost from the other, the money side. And launching is the hard part, where most attempts fail. The book describes three ways to reach critical mass: a zigzag strategy that pushes participation on both sides at once, as YouTube did with uploaders and viewers; a two-step strategy that recruits one side first and uses it to attract the other; and a commitment strategy, for platforms where one side has to invest before it can take part.
Nir Vulkan's The Economics of E-Commerce (Princeton University Press, 2003), written after the first wave of dotcom valuations collapsed, looks at the online marketplace from the design side. It holds that much of e-commerce's value comes from automation: shopping bots and software agents that compare prices, negotiate and bid for their owners, which lower operating costs and create market interactions that did not exist before. Vulkan applies game theory to one-to-one negotiation, one-to-many auctions and many-to-many exchanges, and argues that an online business model has to account for the "total game" its participants play. The design questions this raises are the subject of market design.
Andrew McAfee and Erik Brynjolfsson's Machine, Platform, Crowd (W. W. Norton, 2017) describes three shifts in how businesses organise: from human judgement to machines, from products to platforms, and from a company's own staff to outside crowds. On platforms it follows the economics above. Digital goods cost almost nothing to copy and deliver, which the authors summarise as free, perfect and instant, and a platform grows by admitting complements, as Apple's iPhone did once outside developers could sell apps for it.
The crowd section takes up Bitcoin, blockchains and smart contracts as a way to coordinate economic activity without firms, and its chapter "Are Companies Passé?" answers no. It draws on the concentration of Bitcoin mining and on The DAO, the Ethereum venture fund drained by an attacker in 2016, whose losses were reversed by a hard fork of Ethereum that people rather than the contract decided on (see history of DAOs). The argument is the incomplete-contracts theory of the firm: no contract can specify what everyone will do in every future state of the world, so someone must hold the right to decide what it leaves out, and that residual right is what owning a company means. McAfee and Brynjolfsson expect companies to persist for that reason.
In 2017 blockchain projects argued that a token changes the start-up problem. Michael Karnjanaprakorn, founder of the learning platform Skillshare, set out the case in Token Network Effects (freeCodeCamp, 2017). A network that issues its own token can give early adopters partial ownership of it, and in his account "this new ownership model solves the initial chicken-and-egg problem". He defined token network effects as occurring "when the growth of the network aligns with the appreciation of the token". His examples were Golem, a market for spare computing power, and the Dash, Steem and Kin networks.
Lin William Cong, Ye Li and Neng Wang give the argument a formal model in Tokenomics: Dynamic Adoption and Valuation (The Review of Financial Studies, 2021). Users hold the token to transact on the platform, and because each user's benefit rises with the number of others, adoption follows an S-curve. The token lets users capitalise on the platform's expected growth, which lowers their cost of taking part, and the feedback between adoption and token price speeds adoption and steadies the user base compared with the same platform without a token. The model's pricing results, and the same authors' account of a platform that finances itself by issuing tokens, are covered under token valuation.
The record since 2017 shows where the argument stops. A token that rises with the network pays early users only while it rises, and users who join for the payment tend to leave when it ends. Liquidity mining, in which protocols have paid their own tokens to users since 2020, is the most widely copied form of paying for adoption, and the capital it attracted arrived with the payments and left as they tapered. Token velocity describes a second limit: a token that users pass on as soon as they receive it captures little of the network's value, however large the network grows.
A caper meets the start-up problem on two fronts, and its design removes one. The first is liquidity: a new token usually needs buyers and sellers to arrive together before anyone can trade it. A caper's token trades against a bonding curve at a price set by a fixed formula rather than by matching counterparties, so its market is open from the first buyer and cannot be withdrawn. The price rises as more of the supply is bought, so backers who arrive before demand grows pay less than those who arrive after it, which is the early-adopter stake the token-network argument describes. The second front is demand, and the curve does not remove it: a caper still needs people who want what its members are building.