← all systems · see it in The System
Two-sided market (platform)
A market in which an intermediary, the platform, serves two distinct groups of users who value each other's presence (cardholders and merchants, readers and advertisers, gamers and game developers, riders and drivers), so that the value of joining for one group depends on how many of the other group join (cross-side or indirect network effects). Economists define it by what the platform must choose: not only a price level but a price structure, the split of charges between the two sides, which changes volume and profit because end users cannot pass the charges through to each other (the failure of the Coase theorem is necessary but not sufficient). Platforms therefore often charge one side little, nothing or less than nothing (a loss leader or subsidy side) and recover on the other (the money side). Users may single-home or multihome; where one side single-homes, platforms compete hard for it and charge the multihoming side monopoly prices for access (a competitive bottleneck). Network effects push such markets toward tipping and a few large platforms, so market power, market definition and conduct (interchange fees, anti-steering rules, exclusivity, tying, envelopment) are read differently in competition law, as in Ohio v. American Express (2018) and the EU's card-fee caps and Digital Markets Act. Payment cards, newspapers and media, operating systems and game consoles, and digital platforms (search, marketplaces, ride hailing) are the worked cases.
Properties published scienceour proposal
Sourced fields, and our own readings.
| aliases | Two-sided market |
| born at | per platform: any date; earliest cases read here: newspapers carrying paid advertisements (circulation and advertising markets studied since Corden 1952-53 and Reddaway 1963), the first general-purpose charge card (Diners Club, 1950), BankAmericard (1958); the economics: Rochet-Tirole 2003, Caillaud-Jullien 2003, Armstrong 2006 |
| born precision | n/a |
| boundary | Inside: the platform (the intermediary firm or association that runs the rules, the technology and the prices), the two (or more) groups of users who join it, and the interactions it carries between them (a card transaction, an advertisement seen, a game bought for a console, a ride matched). Outside: the users' alternatives (cash, other platforms they could join or already multihome on, one-sided sellers), the rival platforms, the suppliers of the platform's inputs, and the state and courts that regulate its prices and conduct. Where the line falls is itself contested in law: one market spanning both sides (transaction platforms, per the 2018 majority) or two linked markets (the dissent; media platforms). |
| nested level | Built from a platform firm (or a not-for-profit association of member banks), two user groups and the cross-side network effects between them; a kind of market; embedded in an economy under competition law and the state. Two-sidedness may be a strategy chosen by a firm (Amazon added a marketplace after starting as a retailer) rather than fixed by technology. |
| scale | social |
Measures (31) published science
| Split of a $100 card purchase (US example) | about 98 to the retailer; 1.75 issuer, 0.18 network, 0.07 acquirer US dollars |
| US average interchange rate | 179 basis points (1.79%) |
| US interchange fee, as a share of transaction value | approximately 2 percent |
| EU interchange caps (Regulation 2015/751) | 0.3 credit, 0.2 debit percent of transaction |
| Interchange share of merchants' card fees | 70 to 90 percent |
| Visa and MasterCard interchange fees collected | 26 (2004); 30.7 (2005) billion US dollars |
| Merchant fee to the merchant's bank | typically less than 3 percent of the purchase |
| Card association members | over 6,000 bank and nonbank members (Visa and MasterCard) |
| US credit-card shares by transaction volume (as of 2013) | Visa 45; Amex 26.4; MasterCard 23.3; Discover 5.3 percent |
| Amex merchant price rises without losing a large merchant (2005-2010) | 20 price rises in five years |
| US bottom-quintile households with a credit card | 2 (1970) to over 38 (2001) percent |
| Interchange-fee settlement, US merchants' class action (2013) | 7.25 billion US dollars |
| Waterbed effect in mobile telephony | termination rates down about 10%, mobile retail prices up 5% percent |
| Italian newspapers: observed markup per copy | 0.943 euro per copy |
| Italian newspapers: own price elasticity of readers' demand | around 0.48 elasticity |
| Share of newspaper revenue from advertising | 70-80 percent |
| Share of video game titles released on more than one console | about 12 to almost 40 percent, over 20 years (to about 2005) |
| PlayStation 3 loss per console sold | 250 US dollars per unit |
| Free-PC losses (1999) | 80 million US dollars |
| Ride-hailing surge response period (model) | roughly 5-10 minutes per period |
| Cardholder single-homing spell | months time a consumer sticks to a single card |
| Public awareness of interchange fees (US, January 2007) | about a third of the public had heard of them |
| Reserve Bank of Australia interchange reduction (2003) | about 0.95 to approximately 0.5 percent of the transaction |
| US debit interchange cap (Regulation II) | 21 cents plus 5 basis points of value; up to 1 cent more for fraud prevention (about 24 cents on the average $38 transaction) per debit transaction |
| Amazon's own share of new items listed (books against electronics) | more than 50 (books); closer to 1 (electronics) percent of new items listed sold by Amazon itself |
| Share of eBay buyers who leave feedback | approximately 65 (over 80 in eBay's early days) percent |
| eBay seller percent positive against effective percent positive | percent positive mean 99.3, median 100; EPP mean 64, median 67 percent (about two million sellers, June 2011-May 2014) |
| Annual transfer between cash and card payers (US calibration) | 149 paid by each cash buyer; 1,133 received by each card buyer US dollars a year |
| Banks' card merchant-fee revenue from cash payers | about 79 percent |
| Annual redistribution through credit card rewards (US) | 15 billion US dollars a year |
| UberX consumer surplus (2015) | about 2.9 (four US cities); about 6.8 (US, back-of-the-envelope); about 1.60 per dollar spent billion US dollars |
Open contentions (8) open contention
Where sources or readings disagree. Both stay on the map.
- C-3676 What makes a market two-sided?
booked: The map keeps the competing readings; none is ruled.
the other reading: Rysman: virtually all markets are two-sided to some extent; ask how much it matters; Rochet-Tirole: the cross-group externality definition is endogenous to pricing and can be under-inclusive; Hagiu and Wright 2015: a platform enables direct interactions between distinct sides who keep control of the key terms, each side affiliated with it; indirect network effects neither necessary nor sufficient; supermarkets pass the older tests but are not platforms - C-3677 Are the fees card networks set between banks efficient or anticompetitive?
booked: The map keeps the competing readings; none is ruled.
the other reading: regulated caps (Australia 2003, EU 2015) against the waterbed on cardholder fees and rewards; upheld in the EU: the Court of Justice dismissed MasterCard's appeal (2014); the Commission argued a two-sided market does not need one side to remunerate the other - C-3678 Is a credit-card network one market or two?
booked: The map keeps the competing readings; none is ruled.
the other reading: the dissent: market definition was unnecessary given direct evidence of harm - C-3679 Do the two sides of a platform keep its power in check?
booked: The map keeps the competing readings; none is ruled.
the other reading: anti-steering rules blocked Discover's low-fee route (trial court) against expanding output and quality while the rules were in place (majority) - C-3680 Do platforms tilt their prices toward one side more than is good for society?
booked: The map keeps the competing readings; none is ruled.
the other reading: the two results rest on different settings (usage charges with partial multihoming; membership with one side multihoming), so they may not conflict; Weyl 2010 (abstract): profit maximisation distorts by internalising network externalities only to marginal users (a Spence distortion), so private price structures can be biased; the content of the Rochet-Tirole model turns on the kind of user heterogeneity - C-3681 Who pays for credit-card rewards?
booked: The map keeps the competing readings; none is ruled.
the other reading: reward cards redistribute from naive to sophisticated cardholders regardless of income, about $15 billion a year through interest and fees (Agarwal et al. 2023, Federal Reserve account data); 'reverse Robin Hood' at best incomplete; their abstract also reports the flow running from poorer to richer areas; the two measure different channels (the retail markup on cash payers against interest and fees on reward cards), so they need not conflict; the access side: card holding in the bottom income quintile rose from 2% to over 38% (1970-2001; the Amex majority); a published critique of Schuh-Shy-Stavins (Semeraro 2013) named by the specialist, not opened - C-3683 Were format wars such as VHS against Betamax two-sided platform contests?
booked: The map keeps the competing readings; none is ruled.
the other reading: standards battles are network-effect cases and not cleanly two-sided platforms: a format standard has no sides affiliated with one firm that sets both sides' terms (specialist's reading, leaning on Hagiu and Wright's definition, o-hw-definition) - C-3684 When regulators cap one side of a two-sided market, do prices rise on the other?
booked: The map keeps the competing readings; none is ruled.
the other reading: the same authors' later study (2014, abstract read) finds the waterbed disappeared over 2002-2011 as mobile-to-mobile traffic grew
Cross-scale patterns seen here (22) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Two-sided market (platform). Each links to every other system that shows it.
- conservation law plus closure (structural)
The platform's total price follows the ordinary Lerner rule on the summed elasticities, a balance that leaves the split between the two sides open; a second condition, the ratio of elasticities (and in Ramsey pricing the surplus each side creates for the other), closes it. Under the Coasian closure - runaway from an unstable balance (structural)
Two forms. Tipping: once one platform gains an edge, cross-side effects feed it until it pulls away (a handful of platforms, or one). The demand spiral: a price rise on one side loses members, which lowers the other side's value, which loses more members. The balance point is the platforms' (or side - same role, different builds (structural)
The job is getting two groups who need each other onto one place at prices both will bear. Card networks do it with an interchange fee paid between banks; newspapers with advertising revenue that holds down the cover price; consoles by selling hardware at a loss and charging developers royalties; op - same shape, opposite sign (structural)
One term, the cross-group externality (how much one side's members add to each member of the other side), enters the platform's prices with either sign: positive for cardholders and merchants, negative for readers or viewers facing advertising (ad nuisance). And one pricing rule yields opposite busi - symmetry breaking (name the cut) (structural)
The cut is the price structure: even when the two sides are symmetric, platforms can price them asymmetrically; and platforms that start even can tip to one winner. The symmetric before is shown in the model, the broken after in prices and shares. - trade one property for another (structural)
Network size is bought with network value: the size rule (subsidise the price-sensitive side) raises adoption, the value rule (subsidise those who add platform value) raises price; Adobe and Microsoft settled at different points. Flexibility is paid for with surplus: the side that can multihome gets - who opens the door (structural)
The receiving side builds its door ahead of time: merchants install terminals and pinpads and display decals before any cardholder arrives; the platform posts its fee schedule and rate card first; online debit stalled where merchants had not installed pinpads. A port, opened by the receiver. The sam - borrowed-then-held (partial)
Microsoft: browsers, media players and video editing were first supplied by third parties on Windows, then included in the operating system (borrowed, then held). The reverse also shows: Amazon and Palm first supplied the complement themselves, then opened that side to outside sellers and developers - holding / entrainment (partial)
In a competitive bottleneck the platform sets how many multihoming-side members it carries, to suit itself and its single-homing side; the multihomers' interests are ignored, and too few are served. - nested ladder (partial)
The Coase theorem's failure is necessary but not sufficient for two-sidedness: two-sided markets sit inside the non-Coasian ones. Filistrucchi and co-authors line markets up by pass-through: non-transaction (zero) at one end, transaction platforms (partial) in the middle, one-sided markets (complete - outside-to-inside (partial)
The cross-side externality lies outside each end user's choice (they do not internalise it); the platform takes it inside its own prices. The outside phase is the end users' own trades with no platform pricing the externality; the source states the platform's pricing as the theory's starting point, - same toolkit, independent builds (partial)
Commercial software vendors converged on the same build from the same parts: a free viewer that opens and prints but cannot edit, and a paid writer (Word Viewer, PDF reader, Scientific Viewer). Independence of the builds is asserted by the source's 'converged', not traced. - templating (partial)
One module swapped between systems: the access charge between operators. The Internet's termination charge between backbones is the counterpart of the card associations' interchange fee; mobile termination rates work the same way between networks (the waterbed). The module moved; whether it works th - The carry (partial)
The wait: a ride-hailing match waits for drivers, who respond to a surge much more slowly than riders do; the platform's price settles over periods of roughly 5-10 minutes in the model. The chicken-and-egg start is the same face on a longer clock: one side waits for the other to arrive (RCA had to f - The Payment to Entropy (partial)
Checked against the firing rule. The merchant discount looks like a toll on each card payment ($2 of $100), but each part goes to another holder (issuer $1.75, network $0.18, acquirer $0.07): the dark twin, booked as the exchange with the banks. What fits is the chain form, value lost to stacked mar - the whole sets the local reference (partial)
Each side's price is read against a cost set by the whole: the marginal cost of serving one side is reinterpreted as an opportunity cost, net of what the other side then pays. A price below cost on one side reads as predation locally and as competitive across the whole platform. - Umwelt: each system carves its own slice of one shared medium (partial)
Platforms see different slices of the same interaction between two people: a card network sees every transaction and can charge per use; a newspaper cannot see whether a reader attends to an ad and can charge only for exposure; a ride-hailing platform sees regional demand and supply that its drivers - causal ceiling (surface)
Whether a viewer pays genuine attention to an ad cannot be measured, so attention cannot be priced or compensated; read as a practical limit of monitoring, so surface. - namespace completeness (the map predicts its gaps) (surface)
Rochet and Tirole's Table 2 lists segments 'that will be searching for a proper business model' with the instrument each would use to allocate cost (B2B exchanges, Internet backbones, pools and standards, software as a service), predicted from the theory's grid of subsidised and paying sides; the la - partner-becomes-part (surface)
A stand-alone platform facing envelopment often sells out to the attacker; Microsoft's integration of third-party applications. No gain/gain loop with shared inheritance is shown, so surface. - red-queen (surface)
The rewards race: Amex's rewards drew Visa and MasterCard into premium cards with higher rewards and higher merchant fees, and card pricing has increasingly favoured consumers over time. Both sides improve; the stalemate (no network gaining ground) is not shown in the sources read. - seed (surface)
A platform's volume is the product of its two memberships (n_B times n_S in Rochet and Tirole's volume program): each member who joins one side opens interactions with every member of the other. Read as surface: the sources speak of multiplying pairings, not of an information packet that unpacks new
Valence: what it trades, and with whom (8) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Human (Homo sapiens) via sys:market · Two-sided market (platform) gain
- State (polity); the nation state its predominant modern form via sys:market · Two-sided market (platform) loss
- Money via sys:market · Two-sided market (platform) gain
- Commercial banks (creators of most money) via sys:market · Two-sided market (platform) gain
- Advertisers via sys:market · Two-sided market (platform) gain
- Merchant via sys:market · Two-sided market (platform) gain
- Consumer via sys:market · Two-sided market (platform) gain
- Competition law via sys:market · Two-sided market (platform) mixed
Partners and parts published science
Built from / builds (5)
- Firm (the economic unit)
- Merchant
- Consumer
- Network effect
- Envelopment: an adjacent platform bundling a stand-alone platform's function
Sits inside / holds (3)
expressed by (14)
- Generation/Synthesis
- Templating / Standardization
- Interoperability/Transduction (Bridging)
- Amplification/Scale-Up (Manufacture & Distribution)
- Internal Coordination/Regulation
- Sensing/Scanning
- Trail-Marking / Externalized Memory (Stigmergy)
- Resource Denial/Attrition
- Divide & Isolate
- Reactive Response
- Physical/Virtual Barrier
- Validation of Identity/Intention
- Triage & Prioritized Allocation
- Quality Control & Error Correction
floor filled by (14)
- Human (Homo sapiens)
- State (polity); the nation state its predominant modern form
- Money
- Commercial banks (creators of most money)
- Market and trade
- Monopoly
- Firm (the economic unit)
- Payment card networks (card brands)
- Merchant
- Consumer
- Market competition
- Competition law
- Network effect
- Not-for-profit platform association (card associations of member banks)
Kinds (9)
- Platform (two-sided) dominance
- Market and trade
- Two-sided transaction platform (payment-card type)
- Two-sided non-transaction platform (media type)
- Matchmaker (matching) platform
- Multi-sided platform (three or more sides)
- Not-for-profit platform association (card associations of member banks)
- Shopping mall
- Auction house (traditional and online)
Notes published science
- Your credit card's rewards are paid for partly by people who pay cash: card networks forbid shops to charge card users more, so shops raise prices for everyone, and a Boston Fed calibration found each cash-paying household hands about $149 a year to card users. A later IMF study says the bigger flow runs from less to more financially savvy cardholders, whatever their income (src:schuh-shy-stavins-2010-credit-card, src:agarwal-2023-rewards).
- When European regulators cut what mobile networks could charge for receiving calls by about 10%, customers' prices went up about 5%: squeeze one side of a two-sided market and the other side bulges, the 'waterbed effect'. A decade later the same economists found the bulge had gone (src:genakos-valletti-2011-waterbed, src:genakos-valletti-2014-decade).
- eBay's sellers average 99.3% positive feedback, because unhappy buyers mostly say nothing; count the silences and the average falls to 64% (src:tadelis-2016-reputation).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
the boardroom (shared floors)
- L4: The price structure and level: the two-sided Lerner rule, the split set by elasticities and cross-side effects, the Ramsey benchmark
- L5: Both sides on board: the balance of participation, market power and its check by the cross-side feedback; the platform's governance
- L6: Network scale and platform competition: increasing returns, homing, competitive bottlenecks, tipping or coexistence
- L7: The business model: money side and subsidy side, openness, exclusivity, envelopment
two-sided market as economists describe it: two groups of users who value each other's presence, joined through a platform's prices (reader's view)
- L1: Two groups of end users who value each other's presence: cardholders and merchants, readers and advertisers, gamers and developers, riders and drivers
- L2: Joining: membership on each side (a card in the wallet, a terminal at the till, a subscription, an installed console, an app), each side needing different functionality and its own door
- L3: The interaction carried through the platform (a card transaction, an ad seen by a reader, a game played on a console, a ride matched); the cross-side effect is realised here and not taken into account by the end users
- L4: The price structure and the price level: the total set by a Lerner rule on the summed elasticities, the split by the ratio of elasticities and the cross-side effects; marginal cost read as an opportunity cost
- L5: Getting both sides on board: the balance between the two sides' participation, where the platform's market power sits and where the cross-side feedback checks it; governance (for-profit platform or members' association)
- L6: Network scale and platform competition: increasing returns, single- and multihoming, competitive bottlenecks, tipping to one or a few platforms or coexistence by differentiation
- L7: The business model acted on the market: which side is the money side and which the subsidy side, how open (how many sides, compatibility), exclusivity and steering rules, envelopment; all under competition law and regulation
way in: Competition law and regulators
- L1: Legislatures, courts, competition authorities and central banks (US antitrust, EU competition law, the Reserve Bank of Australia)
- L2: Cases and rules aimed at platforms: interchange caps, no-surcharge and anti-steering cases, gatekeeper rules
- L3: The test the law applies: one market or two, both sides weighed or one, market definition by a two-sided SSNIP
- L4: The remedy acting on a price or a rule: a capped fee on one side, a struck anti-steering clause
way in: Costs of serving each side and each interaction
- L1: The platform's inputs and technology: cards, terminals, consoles, presses, servers
- L2: Fixed costs per member on each side and a marginal cost per interaction
- L3: Marginal cost reinterpreted as an opportunity cost: the cost of one more interaction net of what the other side then pays
way in: Pass-through between end users (whether the two sides can re-price between themselves)
- L1: The end users' own transaction, where a charge on one side could be passed to the other
- L2: Surcharges or separate cash and card prices, blocked by no-surcharge rules, menu costs or unmeasurable attention
- L3: The degree of pass-through: complete (one-sided), partial (transaction platforms) or zero (media)
way in: Rival platforms and multihoming
- L1: Competing platforms (Visa against Amex, consoles of one generation, newspapers against TV)
- L2: Homing: the share of users on each side who single-home or multihome
- L3: Steering: undercutting to make multihomers drop the rival
way in: Side A: participation and demand of the subsidised or single-homing side (cardholders, readers, gamers, users, riders)
- L1: Users who choose whether to join and how much to use
- L2: Membership and usage decisions, each depending on the charges the platform sets
- L3: The side's elasticity, raised by competition and multihoming into an own-brand elasticity
way in: Side B: participation and demand of the money or multihoming side (merchants, advertisers, developers, drivers)
- L1: Merchants, advertisers, developers who decide whether to accept, list or advertise
- L2: Acceptance or listing: the door each seller must install (terminals, pinpads, signage, development kits)
- L3: The side's willingness to pay, which rises with the other side's size
way in: The cross-side effect: how many are on the other side
- L1: The other side's count and activity on the platform
- L2: Demand on one side shifts outward as the other side grows
- L3: The externality the end user does not internalise, priced by the platform as the other side's contribution
way out: Matching and real-time prices (ride hailing, surge pricing)
- L1: Some riders wait out the surge; drivers move into the region
- L3: Riders and drivers matched by the platform
- L4: A surge multiplier set on current demand and supply
way out: Openness: how many sides to run and how compatible to be
- L1: Third-party sellers on Amazon's marketplace; Microsoft's browser inside Windows
- L6: Integrate a side in, or open it to third parties; seek compatibility or incompatibility
- L7: Choose the number of sides and the relation to rival platforms
way out: Rules that bind a side: anti-steering, no-surcharge, honour-all-cards, exclusive contracts
- L1: Merchants cannot steer buyers to cheaper networks; games stay on one console
- L5: Anti-steering and exclusivity clauses written into contracts
- L7: Decide to restrict how a side may treat rival platforms
way out: The money side's price (merchant discount, advertising rates, royalties, developer fees)
- L1: Merchants, advertisers or developers pay for access to the other side
- L4: The fee set on that side, often per transaction
- L7: Decide which side pays
way out: The subsidy side's price: low, free or negative (rewards, free readers, cheap consoles, free papers)
- L1: More members on the subsidised side; the other side's demand shifts out
- L4: A price below cost, zero or negative on that side
- L7: Decide which side is the loss leader