Platform-Mediated Networks: Definitions and Core Concepts PDF

9-807-049 REV: OCTOBER 2, 2007 MODULE NOTE Platform-Mediated Networks: Definitions and Core Concepts Platform-mediated networks encompass users whose interactions are subj...

9-807-049 REV: OCTOBER 2, 2007 MODULE NOTE Platform-Mediated Networks: Definitions and Core Concepts Platform-mediated networks encompass users whose interactions are subject to network effects, along with intermediaries who provide a platform that facilitates users’ interactions. Such networks comprise a large and growing share of the global economy. Examples are as diverse as video games, postal delivery, credit cards, fuel cell-powered cars, instant messaging, web search, real estate brokerage, HMOs, shopping malls, travel reservation systems, DVDs, online dating, and Universal Product Codes. Sixty of the world’s 100 largest companies (ranked by market value) earn most of their revenue from such networks, including American Express, Cisco, Citigroup, Time Warner, UPS, and Vodafone.1 Firms in platform-mediated networks face distinctive management challenges. Familiar rules such as value-based pricing may fail. Traditional barriers to entry may no longer hold. Due to network effects, platform intermediaries often enjoy increasing returns to scale; their industries have room for only a few players. In many platform-mediated markets, network effects are so strong that a single platform prevails. When winners take most, little is left for losers, as evidenced by the spectacular success of some platform providers, for example, Microsoft, eBay, and Google. However, strategic errors are also common when companies strive to create new platforms. Rivals sometimes persist in fighting despite the inevitability of a winner-take-all outcome. The classic example, VHS versus Beta, is being replayed today in a battle over high-definition DVD formats. Likewise, full-frontal assaults in 1999 by Yahoo! and Amazon on eBay’s franchise were expensive failures because strong network effects in online auctions favored a single platform. Firms also miss chances to profitably share their platforms with rivals. For example, many believe that Apple forfeited one of the 20th century’s biggest business opportunities when it failed to license its Macintosh operating system. This note aims to equip students with a vocabulary and frameworks for analyzing platform- mediated networks. It is organized into two sections, plus appendices and a glossary of terms. The first section defines networks and network effects; explains how network effects influence users’ willingness to pay for network access; describes factors that determine the strength of network effects; discusses how a network’s success may depend upon users’ expectations about its growth prospects; defines network externalities and their significance; presents a taxonomy of networks based on the number of distinct user groups—“sides”—that they encompass; and explains why network effects should be viewed as “demand-related” rather than “supply-related” scale economies. ________________________________________________________________________________________________________________ This note was prepared by Professor Thomas R. Eisenmann for the sole purpose of aiding students in Managing Networked Businesses course. Some concepts in the note were developed or refined through conversations with Professors Geoffrey Parker, Marshall Van Alstyne, and Andrei Hagiu of Tulane, Boston University, and Harvard Business School, respectively. Copyright © 2006, 2007 President and Fellows of Harvard College. To order copies or request permission to reproduce materials, call 1-800-545- 7685, write Harvard Business School Publishing, Boston, MA 02163, or go to http://www.hbsp.harvard.edu. No part of this publication may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted in any form or by any means—electronic, mechanical, photocopying, recording, or otherwise—without the permission of Harvard Business School. This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts The second section defines platforms; describes different roles that firms play in platform creation and maintenance; discusses platform boundaries, for example, the distinction between platform providers and network users; and presents schemes for categorizing platforms based on their principal function and one aspect of their structure: who controls them. Networks and Network Effects A network is a system of interconnected nodes. The nodes can be people (e.g., Stanford’s alumni network), companies (e.g., NBC’s network of affiliated TV stations), places (e.g., the network of cities served by British Airways), or things (e.g., computers on a local area network). In platform-mediated networks, the nodes are called network users. Network users are independent actors—individuals and/or firms—who participate in a network in order to interact with each other. This note focuses on platform-mediated business networks that exhibit network effects. Platforms are defined in the next section; briefly, they encompass components and rules required by users to interact through a network. Business networks are organized by profit-seeking firms—intermediaries who create and maintain platforms—to exploit users’ willingness to pay to participate in a network. Network effects are evident when a network’s value to any given user depends upon the number of other users with whom they can interact.2 Delimitations. Our focus on platform-mediated networks excludes: 1) social networks (e.g., a wine- tasting club), unless they are organized by profit-seeking firms; 2) supply networks comprised of a large firm (e.g., Toyota) and its suppliers, unless the firm’s products exhibit network effects; 3) physical networks (e.g., natural gas utilities)—again, unless they distribute products with network effects; and 4) traditional retailers, to the extent that they serve as transaction principals rather than intermediaries. These delimitations are explained in Appendix A. One-, Two-, and Three-Sided Networks Networks can be categorized according to the number of distinct user groups they encompass. In some networks, users are homogenous, that is, they all perform similar functions. For example, although a given stock trade has a buyer and seller, these roles are transient. Almost all stock market participants play both roles at different times. Likewise, almost all telephone network users both originate and receive calls. Networks with homogenous users are called one-sided to distinguish them from two-sided networks, which have two distinct user groups whose respective members consistently play the same role in transactions.3 Examples of two-sided networks include credit cards, comprised of cardholders and merchants; HMOs (patients and doctors); travel reservation services (agents and airlines); and video games (gamers and game developers).4 In a two-sided network, members of each group exhibit a preference regarding the number of users in the other group; these are called cross-side network effects. Each group’s members may also have preferences regarding the number of users in their own group; these are called same-side network effects.5 Figure 1 depicts these relationships. 2 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 Figure 1: Cross-Side and Same-Side Network Effects in a Two-Sided Network Side 1 Side 2 Platform Source: Casewriter. In two-sided networks, users on each side typically require very different functionality from their common platform. In credit card networks, for example, consumers require a unique account, a plastic card, access to phone-based customer service, a monthly bill, etc. Merchants require terminals for authorizing transactions, procedures for submitting charges and receiving payment, “signage” (decals that show the card is accepted), etc. Given these different requirements, platform providers sometimes specialize in serving users on just one side of a two-sided network. Some networks are three-sided: they have three distinct groups of users. Members of each group exhibit a preference regarding the number of users in both other groups. For example, some media platforms (e.g., YouTube) link three sets of network users: content consumers, third-party content providers, and advertisers.6 Network Effects and Willingness to Pay Positive and Negative Network Effects. Network effects influence existing and prospective users’ propensity to participate in a network and their willingness to pay for participation. Network effects can be positive or negative. With positive network effects, a larger user base is appealing. In the classic example, the first fax machine sold was useless until someone purchased a second machine. By providing another potential destination for messages, the subsequent arrival of each new fax machine increased the value of each existing machine—perhaps only by an infinitesimal amount. Similarly, network growth increased the willingness to pay (WTP) of prospects who had not yet acquired a fax machine. With negative network effects, users’ WTP for network participation is reduced as the network’s user base expands. For example, negative network effects eventually became evident in the fax network as its huge user base encountered congestion in the form of busy signals when initiating some transmissions. When a network’s users simultaneously perceive both positive and negative network effects, their relative strength will determine the net impact on WTP. In two-sided networks, cross-side network effects are usually positive, but can be negative (as with consumer reactions to advertising). Same-side network effects may be either positive (e.g., the benefit from swapping video games with more peers) or negative (e.g., the desire to exclude direct rivals from an online business-to-business marketplace). 7 3 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts Willingness to Pay. WTP equals the maximum amount that an individual would pay for a product or service. A product’s aggregate demand curve indicates the total quantity that could be sold at every price level, based on the WTP of each prospective customer. The WTP of network users may be aggregated in the same way to derive a demand curve for network participation. Assuming that network effects are positive, then network growth increases WTP for prospective users. In Figure 2, for example, as the fax network grows from N1 to N2, the aggregate demand curve shifts outward from D1 to D2. Note that all the demand curves pivot at the same point on the horizontal axis, N, which corresponds to a price of zero. When participation is free, everyone who is interested joins the network. WTP and Pricing. When a network’s user base expands, the resulting increase in user WTP does not automatically imply that platform providers will charge higher prices. Competitive dynamics will determine whether increased WTP is captured by providers through higher prices or by users as greater consumer surplus, that is, the sum of the difference between each customer’s WTP and the price actually paid. Even when platform providers have market power due to entry barriers (i.e., the ability to profitably price above marginal cost), they may forego price hikes. Specifically, by sacrificing price increases in the near term, platform providers may be able to build an even bigger network and thereby reap greater profits over the longer term. Likewise, in two-sided networks, a platform provider may profit by permanently subsidizing one side of the network, because growth on that side yields a strong increase in WTP for participants on the other side. Figure 2: WTP and Expected Network Size WTP D3 D2 D1 N1 N2 N3 Ns Network Size Source: Casewriter. Strength of Network Effects This section addresses three questions about the strength of network effects. First, how does the strength of network effects change as networks expand? Second, when comparing different networks, what factors determine the relative strength of their network effects? Third, how can managers measure the strength of network effects? Network Effects and Metcalfe’s Law With positive network effects, users’ WTP for network access increases with network size, but typically at a decreasing rate. This translates into a demand curve that shifts outward by a reduced 4 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 amount with each proportional increase in network size—say, with each doubling, as in Figure 2. Hence, the relationship between a given user’s WTP and network size tends to follow a logistic (“S”- shaped) function. After an initial period of accelerating growth, WTP eventually increases at a decreasing rate due to: 1) budget and attention constraints; and 2) the fact that late adopters conduct fewer transactions and are valued less as transaction partners by existing users. The idea that network effects increase in strength at a diminishing rate with network expansion is at odds with Metcalfe’s Law.8 Robert Metcalfe, who designed the Ethernet protocol, provided the inspiration for Metcalfe’s Law, which states that the value of a network equals the square of the number of its users (“N”). Metcalfe’s Law is based on the properties of mesh networks, in which each network node is linked to every other node. In a mesh network, the number of possible one-way links equals N x (N-1), which approaches N2 when N is large. “N-1” reflects the fact that a node cannot connect to itself. Managers of networked businesses have often invoked Metcalfe’s Law to explain their rapid growth and justify “get-big-fast” strategies.9 However, due to constraints (e.g., I don’t have time to send 70 million faxes) and diminishing benefits (e.g., I do not need to fax strangers in Bolivia), it is almost never true that, as a network grows: 1) per capita transaction volumes also increase in direct proportion with network size; or that 2) each user’s WTP per transaction remains constant. Consequently, Metcalfe’s Law should not be used for business planning. Comparisons across Networks Some networked markets exhibit intrinsically stronger network effects than others. Home buying and mortgage brokerage networks illustrate these dynamics: Home Buying. For both buyers and sellers of homes, maximum marketplace liquidity is crucial. Prospective homebuyers vary widely in their WTP for hundreds of different attributes (e.g., is there a guest bedroom on the first floor for my aging parents? How noisy is street traffic?). Homes for sale in a given area also vary widely in the extent to which they deliver these attributes. If all potential sellers and buyers participate in a single network, then buyers will maximize their odds of finding their perfect house at a fair price, and sellers will maximize their odds of finding the buyer with the highest WTP. Mortgage Brokerage. By contrast, when homebuyers shop for a mortgage using an online broker, they are likely to witness diminishing returns from increased liquidity. Once a buyer specifies a few attributes (e.g., fixed versus floating rate, loan duration), a broker typically delivers quotations from several mortgage issuers. The buyer will appreciate the fact that the service has screened offers from a few dozen issuers; this should be enough to yield a good offer. However, gaining access to thousands of additional quotations is not likely to yield much improvement. Mortgages are commodity products with published prices, offered by huge numbers of banks that compete vigorously. Issuers are unlikely to offer terms that differ substantially from those of their rivals, so there is not much to be gained from affiliating with a platform that offers access to 100% of issuers. Generalizing from these examples, network effects will be strong in matching networks (defined below) when transactions involve heterogeneous demand and supply. If a party on the “ask” side has idiosyncratic requirements, that party will strongly prefer a network that offers access to a wide variety of differentiated partners on the “offer” side. Two other types of transactions are likely to engender strong network effects: 5 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts In some networks, users conduct sequential transactions but require novelty, because repeatedly consuming identical items (e.g., DVDs or video games) would be boring. Some networks facilitate user mobility. For example, credit card holders require access to geographically dispersed stores; adoption for fuel cell-powered cars will hinge on the widespread availability of hydrogen refueling stations. Mobile users value a platform that can accommodate the sheer breadth of locations where they might wish to travel. Measuring Network Effects Network effects are best measured through conjoint analysis, a market research technique in wide use since the early 1970s for designing and pricing new products.10 Conjoint studies elicit consumer reactions to product attributes through paired comparisons (conjoint = CONsidered JOINTly). A common yardstick is used for assessing the value that consumers attach to different attributes, so analysts can predict the impact of changes in pricing, features, and network size on a product’s sales or market share. Conjoint analysis has two potential limitations. First, conjoint studies are expensive, typically ranging from $50,000 to $250,000. Although the technique is a staple of market research, design issues are complex, so input from experienced professionals is necessary. Also, interviewing hundreds or even thousands of consumers is time consuming and costly. Second, conjoint analysis requires that consumers understand product attributes. When a new product has revolutionary features, consumers may have no basis for expressing their preferences, so their survey responses may not reliably predict behavior. The Role of Expectations Users access most networks on a recurring basis over multiple periods (e.g., fax machines, Internet access service). When prospective users decide how much they are willing to pay to participate in a network, the relevant measure of network size is not the current user base, but rather the number of network users with whom the prospect expects to be able to interact in the future. The prospect calculates—at least implicitly—something akin to a net present value. If he or she expects the network’s user base to grow in the future, then WTP should increase for future years. The present value of projected WTP for current and future periods is compared to the present value of expected homing costs. Homing costs include upfront investments and ongoing expenses incurred due to platform affiliation. If the resulting calculation is positive, the prospect should be inclined to join the network. The Penguin Problem In new markets, uncertainty about how technology, business models, and demand will evolve is often high. With fragmented demand—that is, with a very large number of users—it can be difficult for prospective users to communicate their expectations regarding the long-term outlook for a network and coordinate their behavior. Paradoxically, a network with a fragmented base of potential users may stall, even if network effects are strong. I might be willing to pay for network access if I could be assured that many others would, too. Other parties may be in the same position—perhaps enough of them to satisfy mutual requirements for network scale, if everyone actually participated. Due to fragmentation, however, individuals cannot signal their intentions, so they cannot be assured that others will join the network. Facing this uncertainty, each isolated user may defer their purchase. No one moves unless everyone 6 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 moves, so no one moves. (See Appendix B for additional analysis explaining why networked industries are prone to either stall or “tip” toward high penetration.) Economists Joseph Farrell and Garth Saloner labeled this scenario “excess inertia,” and more colloquially, the “penguin problem.”11 Hungry penguins gather at the edge of an ice floe, reluctant to dive into the water. There is food in the water, but a killer whale might be lurking, so no penguin wants to dive first. In such circumstances, individual rationality may lead a group to forfeit attractive opportunities, for example, a predator-free meal or an innovative new networked product. Externalities Excess inertia is caused by network externalities. An externality is a benefit or harm experienced by party B due to the actions of party A, with no compensating payment between the parties. In the case of a positive externality, B pays no price for a benefit provided by A (e.g., B admiring flowers planted in A’s yard). With negative externalities, B receives no recompense for harm caused by A (e.g., B suffering the stench from A’s upwind pigpen). The fax market’s network effects are positive externalities. Owners of existing machines make no payments for the extra value they receive when a new machine is purchased. This positive externality is due in part to the fragmented structure of the mature fax market. Today, adding one more machine to the fax network increases the value received by each existing machine owner by a minuscule amount. However, this tiny increase in WTP, multiplied by 70 million—the current global installed base of fax machines—might well represent a meaningful fraction of a machine’s purchase price (see Appendix C, which estimates this fraction for spreadsheet software). The owner of the new machine receives no compensation for benefits she creates for other network users. Even if platform providers—in this case, scores of firms that sell fax machines—were able and willing to coordinate cross-user payments, there is no practical way to track or transfer such tiny value increments. Internalizing Externalities Absent a way to “internalize externalities,” that is, to compensate (or charge) new users for the incremental benefits (or harm) they bring to the rest of the network, prospects are less likely to join a network, and its growth could conceivably stall. Depending on the network’s structure, however, participants may be able to internalize the externality in three ways.12 Side Payments. Users may be able to “bribe” otherwise reluctant partners. For example, a grandmother may be unwilling to buy a videophone, but her children may find it worthwhile to buy her one. Side payments are more viable with concentrated network usage patterns, that is, when users complete a large share of their transactions with a small number of partners. Subsidies for Early Adopters. A platform provider may offer discounts or promotional inducements to early adopters in order to “prime the pump.” This is a two-phased gambit. The platform provider’s bet is that by accelerating the network’s early growth during phase one, other prospects will revise upward their expectations regarding its potential scale. These prospects will then be more inclined to join, and network growth will snowball. As later adopters’ expectations about the eventual size of the network ratchet up, so does their willingness to pay. If the platform provider can exploit this increased WTP by charging a higher price during phase two, then phase two profits may be big enough to offset phase one losses. In this manner, the platform provider internalizes the network externality inter-temporally. In effect, early adopters who might not otherwise join the fledgling network receive a subsidy that compensates them for the value they ultimately will bring to late adopters. This strategy is risky, 7 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts however, since there is no guarantee that phase two customers will sign up. Furthermore, a proprietary platform—one with a single provider—is typically required for this strategy, because the same provider that makes phase one investments must be able to reap phase two profits. If multiple platform providers share a network, provider A might make phase one investments, only to see providers B and C exploit these investments as free riders during phase two. Permanent Subsidies in Two-Sided Networks. In two-sided networks, a proprietary platform provider may permanently subsidize users on one side. These users (for convenience, “side A”) are charged less than they would pay if their platform provider ignored cross-side network effects and priced platform access as if the “side” was an independent market. This subsidy expands Side A’s user base, which in turn increases side B users’ willingness to pay for network access, due to positive cross-side network effects. The platform provider then boosts fees to side B (the “money side”), extracting more than enough additional rent to recover profit foregone on side A (the “subsidy side”). The subsidy to a new side A user internalizes the externality engendered by her network participation, that is, the incremental benefits realized by side B users from interacting with an additional side A user. Permanent subsidies are common in two-sided networks. Examples include pricing video game consoles below their manufacturing cost, granting job seekers free access to online recruitment sites, and shoppers receiving free parking at malls. As with subsidies for early adopters, permanent subsidies are generally viable only for proprietary platform providers. When multiple providers share a platform, free riders may seek to profit by withholding subsidies and serving only money side users.13 Demand- vs. Supply-Related Scale Economies Companies enjoy scale economies when their profit margins improve with expanding sales volume. Network effects are “demand-related” scale economies because they influence the revenue side of a platform provider’s profit equation through their impact on user adoption and willingness to pay. “Supply-related” scale economies are realized when firms reduce unit costs either by spreading fixed expenses over a bigger sales volume or by leveraging experience effects, that is, “learning by doing.” Platform intermediaries sometimes enjoy strong supply-related scale economies because they require heavy upfront investments that yield fixed-cost leverage. For example, launching a new video game console entails huge investments in custom semiconductor design. Likewise, many networked businesses—for example, software development—offer potential for learning by doing because they involve complex, multi-step production processes that lend themselves to ongoing improvement. Demand- and supply-related scale economies are conceptually distinct and should not be conflated. Unit cost reductions that result from network growth should not be labeled “network effects.” Of course, it is crucial for managers to be cognizant of both types of scale economies when formulating strategy. Platforms In the context of product design, platforms encompass the subset of components used in common across multiple offerings in a product family.14 For example, Chrysler’s K-car platform of the 1980s shared a chassis, drive train, and engine across several models. In the context of networked markets, the definition of a platform is similar. In a platform-mediated network, a platform encompasses the common components and rules employed by network users in most of their interactions.15 “Components” include hardware, software, and services required by users, along with an architecture 8 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 that specifies how these components fit together. “Rules” include standards that ensure technical compatibility between components, protocols that govern information exchange (e.g., authentication procedures), policies constraining the behavior of network users (e.g., whether copyrighted content may be shared), and contracts specifying terms of exchange and the rights and responsibilities of network participants. For example, from an Xbox gamer’s perspective, the Xbox platform includes a console and its user interface software; these are required components for each Xbox-compatible game he plays. On the other side of the Xbox network, game developers use a different set of components and rules across multiple projects. For developers, the Xbox platform is comprised of software development kits (SDKs), guidelines for game design, and legal contracts that cover royalties and quality-assurance arrangements—all provided by Xbox’s owner, Microsoft. eBay is another example of a platform. To use eBay, bidders and sellers—the network users— must have access to an Internet-connected PC with a web browser. Other platform components include eBay’s web pages, its bid tracking software, shipping services, and payment services (e.g., PayPal, a separate platform owned by eBay). Platform rules include registration requirements, policies for bidding (e.g., how long auctions last), price terms, dispute resolution procedures, and feedback systems for tracking sellers’ trustworthiness. Note that platform intermediaries do not necessarily own or control all platform components. eBay, for example, does not provide its users with a browser or shipping services. Platform Intermediary Roles Interactions in platform-mediated networks always entail a triangular pattern of relationships. To interact with each other, two users access a common platform—the third point on the triangle.16 Network users will choose to rely on platforms when doing so improves transaction efficiency or effectiveness, compared to unmediated dealings. Platforms are not necessarily created and maintained by a single firm. We distinguish between three roles: platform provider, platform sponsor, and platform component supplier. Platform providers mediate network users’ interactions; they serve as users’ primary point of contact with the platform. Platform sponsors do not deal directly with network users; rather, they hold rights to modify the platform’s technology and to determine who may participate in the network as platform providers, component suppliers, and network users. As noted below, for some platforms these rights are diffused or shared so widely that no single entity serves as sponsor. Platform component suppliers make available to network users essential platform goods and services that are not offered directly by platform providers. For a given platform, the sponsor and provider roles each may be filled by one company or shared by multiple firms; often, a single firm plays both roles (e.g., eBay, Monster.com, Xbox). Identifying Platform Providers. When several parties supply different platform components, we can determine which is the platform provider by asking: Who does a user contact first when a transaction goes awry? For example, satellite radio users require a subscription from a service provider—XM or Sirius—along with a radio. For an “aftermarket” installation (i.e., adding a satellite radio receiver to vehicle already on the road, rather than acquiring a brand-new car with factory- installed satellite radio), the user can choose between many radio brands whose manufacturers are 9 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts licensed by XM and/or Sirius to supply compatible hardware. If they have a reception problem, however, users are likely to initially contact XM or Sirius—their platform provider—rather than their radio’s manufacturer. Component-Based Network Effects. Platform components may be subject to a third type of network effect, distinct from the same- and cross-side effects described above: component-based network effects. With such effects, network users exhibit preferences for variety in their choice of platform components; their willingness to pay for platform affiliation increases when they have access to a larger number of suppliers. For example, prospective subscribers will find XM’s platform more attractive if a wide selection of XM-compatible radios is available (e.g., in-car models, boom- boxes, handheld units). Likewise, radio manufacturers will be willing to pay more for a license to produce XM-compatible hardware if the platform attracts large numbers of subscribers. Platform Sponsors Sponsored vs. Unsponsored Platforms. Platforms may be sponsored or unsponsored.17 Some platforms have no obvious sponsor; they evolve through the collective efforts of platform providers, component suppliers, network users, government regulators, standards-setting organizations (SSOs), and other parties. The Internet is an example of an unsponsored platform. Responsibility for improving the Internet’s core technologies is spread across several SSOs, most notably the Internet Engineering Task Force (IETF) and the World Wide Web Consortium (W3C). National governments typically hold the right to decide who may serve as a platform provider, that is, an Internet Service Provider (ISP). Other examples of unsponsored platforms include the electric power grid and internal combustion engine-powered transportation. One vs. Many Sponsors. Sponsored platforms may have a sole sponsor, that is, a single firm that holds the rights to change platform technology and designate network participants. Examples of sole platform sponsorship include eBay, Skype’s Voice over Internet Protocol (VoIP) network, the American Express credit card, and Apple’s Macintosh operating system. Alternatively, multiple parties may jointly sponsor a platform, typically under the auspices of some association with formal procedures for securing agreement among its members. For example, the Institute for Electrical and Electronics Engineers (IEEE) Standards Association, which has about 8,000 individual and 60 corporate members, sets standards for wireless local area network (WLAN) technologies (e.g., 802.11g). Likewise, Visa is an association jointly sponsored by 21,000 member banks. One vs. Many Platform Providers Sole platform sponsors may serve as a platform’s only provider. Examples of such proprietary platforms include eBay or Microsoft’s Xbox. Sometimes, however, a sole sponsor may license other platform providers. For example, starting in 2004, American Express licensed permission to issue American Express-branded credit cards to third-party banks such as MBNA. Likewise, Apple briefly licensed its Macintosh operating system to “clone makers” in the mid-1990s, but terminated these relationships and once again serves as sole provider of the Macintosh platform. Joint platform sponsorship usually results in a shared platform served by multiple platform providers. For example, the DVD standard was the collective invention of ten companies (Hitachi, Philips, Matsushita, Mitsubishi, JVC, Pioneer, Sony, Thomson, Time Warner, and Toshiba) that pooled technologies in 1995. These platform sponsors jointly own the DVD Format/Logo Licensing Corporation, which in turn permits hundreds of other manufacturers—platform providers—to use the DVD format, as long as they pay license fees to the Corporation and meet its quality standards. 10 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 In two-sided networks, different platform providers may serve the two users who complete a network transaction. For example, the two parties to a Visa credit card transaction—the cardholder and merchant—are likely to be served by different platform providers—two different banks, the “issuer” and “acquirer,” respectively—both of whom are members of the association that serves as the Visa platform’s joint sponsor. Figure 3 illustrates network relationships for Visa, a jointly- sponsored platform with multiple providers, versus Xbox, which has a sole sponsor that also serves as its platform’s only provider. Figure 3: Comparison of Visa and Xbox Platforms VISA = XBOX = JOINTLY-SPONSORED SOLE-SPONSORED SHARED PLATFORM PROPRIETARY PLATFOR CARD HOLDERS MERCHANTS GAMER DEVELOPER Users ISSUING ACQUIRING CONSOLE SDK Platform BANKS BANKS Providers MICROSOFT XBOX Platform VISA INTERNATIONAL Sponsors Source: Casewriter. Compatibility and Interoperability Shared platforms providers employ differentiated but compatible technologies: any network user can switch providers and still interact with the same partners as before (e.g., the owner of a Compaq Windows-compatible PC can switch to a Dell Windows-compatible PC and still use the same applications and peripherals). By contrast, rival platforms employ non-compatible technologies (e.g., Playstation versus Xbox, Visa versus American Express; Windows versus Macintosh). Absent technical modifications that facilitate interoperability, rival platforms’ respective user bases cannot interact. Such modifications may result from mutual consent. For example, Microsoft and Yahoo! recently agreed to interconnect their previously proprietary instant messaging networks. Alternatively, modifications that facilitate interoperability may result from one platform’s unilateral actions. For example, without Apple’s permission, RealNetworks offered software that permitted iPod owners to use RealNetwork’s music store instead of Apple’s iTunes. “Open” vs. “Closed” Platform-Mediated Networks One function of platform sponsors is determining who may participate in a network in the role of platform provider or network user. When no restrictions are placed on participation, we refer to a role as being “open.” 11 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts The rapid growth of open source software (e.g., Linux) and content created through collaborative user communities (e.g., Wikipedia; virtual worlds like Second Life) has generated enthusiasm about the merits of open versus closed network models. Describing a platform-mediated network as “open” can cause confusion, however, because networks have several sets of participants playing different roles. For a given network, participation within one role may be open while participation in the others remains restricted, that is, “closed.” To avoid confusion, whenever using the term “open,” we should specify which roles are being referenced. For example, the 802.11 (“Wi-Fi”) network is unambiguously open for all network participants. With respect to platform sponsorship, anyone can join the IEEE and participate in its standards- setting processes. Hundreds of firms can and do serve as platform providers, because 802.11 technologies are available to all interested manufacturers under “reasonable and non-discriminatory” (RAND) licensing terms, which apply for all IEEE standards. Finally, because anyone can buy 802.11- compliant equipment, participation is fully open for both groups of potential network users: individuals who connect to the Internet using 802.11 cards in laptops and other devices, and organizations or households who deploy 802.11 access points. By contrast, for the Xbox network, Microsoft is the platform’s sole sponsor and its proprietary provider. These roles are closed: no-one but Microsoft can modify Xbox technology or manufacture Xbox consoles. Xbox has two distinct groups of network users: game players and developers. Network access is open for players: anyone can purchase an Xbox console. However, to ensure quality, Microsoft imposes strict licensing requirements on game developers. A developer can only publish an Xbox-compatible game with Microsoft’s permission. Table 1 shows how several platforms vary in the extent to which they are open. Explaining factors that influence a platform’s optimal degree of openness is beyond the scope of this note, but Table 1 suggests that platforms can succeed with very different degrees of “openness.” Table 1: Comparison of Degree of Openness Across Platforms Linux Windows Mac Xbox User Side #1 Open Open Open Open End User User Side #2 Open Open Open Closed Developer Platform Provider Open Open Closed Closed (hardware/operating system bundle) Platform Sponsor Open Closed Closed Closed Source: Casewriter. Platform Boundaries Every platform-mediated network has one, and only one, platform at its core. Every platform facilitates user interactions on one, and only one, network. In most contexts, we can use the terms “network” and “platform” interchangeably, without risking confusion. For example, we mean the same thing when we say that an individual uses the eBay network or the eBay platform. To be precise, however, a network is comprised of all its users, its platform provider(s), its platform sponsor (if one exists), and any platform component suppliers. 12 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 An individual firm may simultaneously serve as sponsor and/or provider for several different platforms. DoCoMo, for example, is a mobile phone service provider (an unsponsored shared platform) that offers i-mode mobile data service (a sole-sponsored proprietary platform) and has recently launched Mobile FeliCa payment services (a jointly-sponsored shared platform).18 Although DoCoMo leverages many common components and user relationships across these three platforms, the platforms and the networks they serve are distinct. The roles of “network user” and “platform provider” are not fixed and absolute. An entity may participate simultaneously in multiple networks; that entity’s role in any given network is defined by the nature of its interactions with other network participants. For example, eBay serves as a platform provider in the context of one network—online auctions—and as a network user in the context of another network—the World Wide Web. Nested Platforms. Extending the eBay example, we see that platforms are often nested within other platforms, like Russian matryoshka dolls. For example, the public switched telephone network (platform #1) provides digital subscriber line and dial-up connections used to access the Internet (platform #2). Internet users also require some type of terminal, such as a personal computer, which by necessity incorporates an operating system, for example, Windows (platform #3). The Internet supports many information transfer protocols, including the World Wide Web’s hypertext transfer protocol (http) and http-compliant browsers and servers (platform #4). Many web sites, like eBay, serve as platforms themselves (platform #5) and may provide access to still more platforms, like the payment processing service PayPal (platform #6). Integration into Network User Role. Platform providers are intermediaries; they are not themselves network users unless they have integrated across both roles, as with Microsoft’s Xbox business. As noted above, Microsoft serves consumers and third-party game developers as Xbox’s platform provider. However, as a “first-party” developer of some Xbox-compatible games like Halo, Microsoft also plays the role of network user. When Are Components Part of the Platform? Consider the huge array of hardware, software, and service components that can be used with a Windows-compatible personal computer: browsers, scanners, Voice-over-IP (VoIP) software, password management utilities, Internet access service, modems, word processors, monitors, CD burners, printers, web-based email, screen savers, keyboards, etc. When should a given component be considered: a) Part of the Windows PC platform, supplied by either platform providers (i.e., PC makers like Dell) or by platform component suppliers? versus b) A complement to the platform that is supplied by a network user? Based on our definition of a platform—the subset of components and rules employed by users across a large majority of their network transactions—components are part of a platform when they are nearly indispensable for most users on at least one side of a multi-sided network. For example, with rare exceptions, PC users must have a monitor and a keyboard, so these components are part of the PC platform. By contrast, only a modest fraction of PC users require scanners or VoIP software, so these components are complements supplied by network users. However, platforms are not static; they evolve over time. Hence, a component’s status may change as a greater share of users comes to view it as indispensable. For example, Microsoft has absorbed many components into the Windows platform that previously were complements supplied by third-party network users, including its web browser, Wi-Fi software, and audio/video players. 13 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts Categorizing Platforms Platforms can be categorized in two ways, based on: Their structure; specifically, whether the sponsor and platform provider roles are respectively controlled by a single firm or shared by multiple parties. The function they fulfill for network users: providing point-to-point connectivity, promulgating variety, matching, or price setting. Structure of Platform Control As noted above, a platform’s sponsor and provider roles each may be filled by one company or may be shared by multiple firms; often, a single company fills both roles. These possibilities define four different types of platforms, shown in Table 2. A proprietary platform has a single sponsor who serves as its sole provider, for example, Apple Macintosh or Sony Playstation. With a shared platform such as Visa or Linux, multiple firms collaborate in developing the platform’s technology and then compete with each other in providing differentiated but compatible versions of the platform. With a joint venture model, several firms cooperate in developing the platform, but a single entity serves as its sole provider (e.g., the online recruitment site CareerBuilder, created by three newspaper groups). With a licensor model, a single company develops a platform’s technology then licenses other providers (e.g., American Express, which licensed MBNA to issue its cards in 2004). A fundamental design decision for firms that aspire to create platform-mediated networks is whether to preserve proprietary control of the new platform’s sponsor and provider roles, or share them with partners. Analysis of factors favoring different models is beyond this note’s scope.19 Table 2: Examples Categorized Using Structural Taxonomy of Platforms One Provider Many Providers Proprietary Licensor One Apple Macintosh American Express (licensed MBNA, etc.) Sponsor Monster.com Palm OS (licensed Sony, Samsung, etc.) Sony Playstation Scientific-Atlanta (licensed “clone makers”) Joint Venture Shared Many CareerBuilder (created by 3 newspaper groups) Linux Sponsors Covisint (created by big auto makers) Real estate multiple listing service (MLS) Orbitz (created by several airlines) Visa Source: Casewriter. 14 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 Platform Function Platforms also can be categorized into four groups based on their principal function: providing connectivity, promulgating variety, matching, and price setting. As shown in Table 3, the four functional types can be distinguished along two dimensions: 1) whether they are one- or multi-sided; and 2) whether parties to a transaction typically rely on their common platform to discover each other or know each other’s identity at the transaction’s outset.20 Connectivity. Connectivity platforms facilitate point-to-point transfers of information, goods, or passengers between two parties or locations (e.g., email; package delivery). To initiate a transfer, a connectivity network user must know their transaction partner’s identity—that is, their “address.” Connectivity networks usually have one-sided structures; users alternate between the roles of sender and receiver.21 Variety. Variety platforms invariably serve multi-sided networks.22 They elicit offers from supply-side users that vary along dimensions valued by demand-side users (e.g., novelty, as with video games or movies; location convenience, as with ATMs or Wi-Fi access points).23 Supply-side offerings are often complements to a platform good possessed by demand-side users, for example, video games vis-à-vis consoles. These complements and platform goods must be consumed in tandem; typically, neither has value without the other. Before they can access a complement offered by a supply-side user, a demand-side user must know the complement’s identity. Subsequent transactions are likely to involve different complements, because demand-side users (often called “end users”) have a preference for variety. For example, they would not wish to play the same video game repeatedly or be restricted to a single ATM location.24 Matching. Matching platforms invariably serve two-sided networks. Platform providers facilitate a process through which members of two distinct groups of network users with heterogeneous needs and offers discover suitable transaction partners in the opposite group.25 Examples of matching platforms include online recruitment sites, real estate brokerage firms, and online dating services. Price-Setting. As with matching platforms, users rely on price-setting platforms to find transaction partners. Whereas matching platforms help users reveal idiosyncratic requirements and offers, users in price-setting networks disclose the price at which they are willing to exchange well-defined items, for example, a share of General Electric stock. Examples of price-setting networks include securities exchanges, auctions, and gambling bookmakers.26 Price-setting networks usually have one-sided structures because they often include traders, who regularly alternate between the roles of buyer and seller to exploit arbitrage opportunities. 15 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts Table 3: Examples Categorized Using Functional Taxonomy of Platforms Transaction Partner’s Identity is Generally: Known Ex Ante Discovered Ex Post Connectivity Platforms: facilitate point-to-point transfers Price-Setting Platforms: permit users to disclose prices at between two parties or locations which they are willing to exchange well-defined items Phone, fax, email, postal service, instant messaging, Securities and commodity exchanges One- text messaging, Bluetooth, CB radio, etc. Auctions Sided Package delivery Wholesale electric power markets Airlines, railroads, etc. Bookmakers Variety Platforms: elicit supply-side offers that vary along Matching Platforms: help users with heterogeneous dimensions valued by the demand-side needs/offers find suitable transaction partners DVD, video game consoles, CD, HDTV, World Wide Real estate brokerage Web, satellite radio, etc. Travel agencies Online media players (e.g., PDF, Flash) Online car buying services PCs, smart phone and PDA operating systems Insurance brokerage Two- Ad-supported media (e.g., magazines, TV) Executive search, online recruiting, talent agencies Sided Credit cards, ATMs, Wi-Fi, gasoline refueling stations Yellow Pages, paid search (e.g., Google) Container shipping Online dating, personal ads, marriage brokers, Barcodes, RFID nightclubs, brothels, etc. Shopping malls HMOs Source: Casewriter. 16 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 Glossary Business networks: Organized by profit-seeking firms—intermediaries who create and maintain platforms—in order to exploit users’ willingness-to-pay for access to a network. Some business networks leverage social networks, but most social networks are not organized by profit-seeking firms. Closed network: A network with access restrictions, e.g., Xbox or Macintosh. Platform sponsors may deny some parties the right to serve as platform providers or network users. It is important to specify which network roles are closed. Contrast with open networks. Compatibility: Technical and/or contractual specifications that ensure that network users can interact, despite being affiliated with rival providers of a shared platform. For example, any Windows- compatible application will run on any Windows-compatible PC. Distinguish from interoperability, which involves rival platforms. Complement: A good that gains value when consumed in tandem with another good (e.g., razors and blades). Complements exhibit negative cross-price elasticity (i.e., a price increase for one good leads to a reduction in the other’s sales volume), in contrast to substitutes, which have positive cross- price elasticity. Component-based network effect: Network users derive value as additional platform intermediaries—either platform providers or platform component suppliers—offer variants of a given platform component (e.g., different PC brands for Windows-compatible application users; different payment processing services for online auction users). Likewise, intermediaries who offer a variant of the component derive greater value from platform affiliation as the network’s user base expands. Distinguish from same-side and cross-side network effects, which involve transactions between network users. Connectivity platform: Facilitates point-to-point transfers—of information, physical goods, or passengers—between two parties or locations. Examples are prevalent in telecommunications (e.g., email, phone, SMS, IM) and transportation (e.g., package delivery, passenger airlines). One of four types in a taxonomy of based on platforms’ principal function. Conjoint analysis: Market research technique that uses paired comparisons to estimate consumers’ preferences for various product attributes, including features and price. Can be used to measure network effects. Cross-side network effect: A preference by users for more or fewer users on another side of a multi- sided network. Distinguish from same-side network effects. Demand-related scale economies: Economies of scale related to customer demand. Positive network effects are demand-related economies of scale, because users’ willingness-to-pay for network access and their propensity to participate in the network increases with growth in the network’s user base. Distinguish from supply-related scale economies. Demand-side users: In a variety network, users who demand a variety of goods or services from users on the supply-side. Ecosystem: See supply network. 17 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts Excess inertia: Also known as the penguin problem. The propensity for a network to stall, despite positive network effects, because: 1) a fragmented population of prospective users is unable to communicate expectations and coordinate behavior; and 2) no proprietary platform provider exists that can internalize network externalities, that is, “prime the pump” through investments that stimulate early adoption. See also excess momentum. Excess momentum: Also known as the lemming effect. The propensity for a new network to rapidly displace an older one, notwithstanding the fact that most of the old network’s users would prefer to avoid incurring switching costs. Fragmentation prevents the old network’s users from communicating their preferences and coordinating behavior; many users infer—incorrectly—that most other users prefer to switch, and all wish to avoid being stranded in a sub-scale network. Externality: A benefit or cost borne by one party due to another party’s actions, with no compensating payment between the parties. Fragmentation: A market’s buy side or sell side is fragmented when it is comprised of large numbers of participants, none of whom have transaction volumes big enough to influence market price. Fragmentation is significant in networked markets because it can lead to excess inertia and excess momentum. Homing: A synonym for network user’s affiliation with a platform. Homing costs: Upfront, ongoing, and termination costs borne by users due to platform affiliation. Should be distinguished from switching costs. Internalize externalities: Provide compensating payment and thereby neutralize an externality. When network users are large or have just a few preferred transaction partners, they can internalize prospective partners’ network externalities by providing side payments to encourage their adoption. A proprietary platform provider can internalize network externalities by subsidizing certain users, catalyzing positive feedback fueled by network effects. After more users join, the platform provider can charge fees to some or all users that recover the subsidies. Interoperability: Technical and/or contractual specifications that ensure that users of two different platforms can interact (e.g., Microsoft and Yahoo! allowing users of their previously proprietary IM services to communicate). Distinguish from compatibility, which involves rival providers of a shared platform. Joint platform sponsorship: Multiple parties hold platform sponsorship rights, often under the auspices of an association with formal procedures for securing agreement among the parties, for example, a standards-setting body like the IEEE. Joint venture platform. A platform with multiple sponsors but one platform provider. One of four types in a taxonomy of platforms based on control structure. Lemming effect: See excess momentum. Licensor platform: A platform with a single sponsor that licenses other platform providers. One of four types in a taxonomy of platforms based on control structure. Matching platform: Facilitates a discovery process that helps network users with heterogeneous needs and offers screen pools of potential transaction partners seeking new relationships and/or attractive trading opportunities. Examples include real estate brokers, online dating sites, and executive search firms. One of four types in a taxonomy of based on platforms’ principal function. 18 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 Metcalfe’s Law: The assertion—almost never valid for real-world platform-mediated networks—that the value of a network equals the square of the number of its users. Named for Robert Metcalfe, the inventor of Ethernet technology, Metcalfe’s Law is based on the number of possible connections in a mesh network, in which every node is connected to every other node. Multi-sided network: In multi-sided networks, groups of users permanently play distinctly different roles in transactions (e.g., job seekers and recruiters; gamers and game developers). Most multi-sided networks are two-sided; some are three-sided. Contrast with one-sided networks. Negative network effect: Users prefer the ability to interact with fewer other users on the relevant side of a network, for example, due to congestion in connectivity networks. Contrast with a positive network effect. Nested platforms: An entity is nested when simultaneously it is one network’s user and another’s platform. For example, eBay is a site on the World Wide Web (a user of that platform) and at the same time a platform that facilitates user access to an online auction network. Network: A system of interconnected nodes. In the context of platform-mediated networks, a network is comprised of network users who wish to interact, plus a platform that facilitates interactions. Networks are served by one, and only one, platform; each platform serves only one network. Network effects: Evident when a network’s value to any given user depends on the number of other users with whom they can interact. Network effects may be positive or negative and may be same-sided, cross-sided, or component-based. Network effects are demand-related scale economies. Network externality: An externality due to a network effect. Network user: An independent actor—an individual or organization—that transacts with other users through a common platform. Networked market: Two platforms serve users in the same networked market when they are substitutes, specifically, when changing one platform’s homing costs influences transaction volumes for the other platform. One-sided network: In one-sided networks, users play transient rather than permanent roles in transactions (e.g., a stock buyer today may be a seller tomorrow). Contrast with multi-sided networks. Open network: A network with no access restrictions imposed by its platform sponsor. It is important to specify which network roles are “open.” Contrast with a closed network. Penguin problem: See excess inertia. Physical network: Physical infrastructure with a networked architecture (e.g., cable TV systems, natural gas pipelines). Some, but not all, businesses that deploy physical networks also exhibit network effects. Platform: In the context of product design, the subset of components used in common across multiple offerings in a product family. In the context of platform-mediated networks, platforms encompass the subset of components and rules employed by users across a large majority of their network transactions. Platforms serve one, and only one, network; each network is served by just one platform. Platforms may be sponsored or unsponsored. Depending on whether their sponsor and provider roles are each controlled by one or many firms, platforms may employ proprietary, shared, 19 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts joint venture, or licensor control structures. Their principal function may be providing connectivity, promulgating variety, matching, or price setting. Platform component supplier: Makes available essential platform goods and services that are not offered directly to network users by platform providers. For example, shipping services are component suppliers for eBay’s platform. Platform-mediated networks: Networks with a triangular pattern of relationships in which two parties to a transaction—network users—each access a common platform that facilitates their transaction. Platform provider: An entity that mediates network users’ interactions and serves as their primary point of contact with a platform. Platform sponsor: Entity holding clear rights to specify who can modify the platform’s technology and who can participate in a network. Sponsors do not deal directly with network users. Sponsorship may be sole or joint; some platforms are unsponsored. Positive network effect: Users prefer the ability to interact with greater numbers of other users on the relevant side of a network. Contrast with a negative network effect. Price-setting platform: Facilitates transactions between users with different willingness-to-pay for well-defined items, such as a share of stock. One of four types in a taxonomy of based on platforms’ principal function. Proprietary platform: A platform with a single sponsor that also serves as its sole platform provider (e.g., Xbox, eBay). Contrast with a shared platform. One of four types in a taxonomy of platforms based on control structure. Same-side network effect: A preference by users for more or fewer users on their own side of a multi-sided network. Distinguish from cross-side network effects. Shared platform: A platform with multiple sponsors and multiple competing providers who rely on compatible components. One of four types in a taxonomy of platforms based on control structure. Side: In the context of multi-sided networks, a side is a group of users who consistently play the same role in transactions. For example, job seekers and recruiters comprise the two sides of online recruiting networks. Social network: A set of dyadic relationships between individuals, reflecting attributes such as frequency of interaction (who talks to whom?), power (who influences whom?), and affinity (who likes whom?). Some platform-mediated networks leverage users’ social networks (e.g., instant messaging services), but most social networks are not business networks. Sole platform sponsorship: A single firm that holds platform sponsorship rights. Contrast with joint platform sponsorship. Sponsored network: A network with a clear platform sponsor. Contrast with unsponsored networks. Supply network: Consists of a large firm (e.g., Toyota) and all its suppliers. Also known as an ecosystem. Supply networks are platform-mediated networks only if they exhibit network effects. 20 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 Supply-related scale economies: Unit cost reduction associated with growth in production volume, due to either fixed cost leverage or experience effects (i.e., “learning by doing”). Unit cost reductions that result from network growth are supply-related economies; they should not be labeled network effects, which are strictly demand-related scale economies. Supply-side users: In a variety network, users who offer a variety of goods or services required by users on the demand side. Switching cost: Out-of-pocket expenses and inconveniences incurred by network users (or by platform providers on their behalf) when users replace one platform provider with another. Distinguish from homing costs. Three-sided network: Has three distinct user groups, each of which demonstrates a preference regarding the number of users in the other two groups. Examples include some media platforms that serve consumers, third-party content providers, and advertisers (e.g., YouTube). Two-sided market: A networked market with two distinct sets of users. Networked markets may incorporate multiple rival platforms. Consequently, a two-sided market (e.g., the credit card industry) may encompass multiple two-sided networks (e.g., Visa, American Express, etc.). Two-sided network: A network with two separate sets of users that interact with each other (e.g., game players and developers, job seekers and recruiters). Roles are stable in two-sided networks; users on one side rarely migrate to the other side. Contrast with one-sided networks, and distinguish from two-sided markets. Two-sided platform: Serves a specific two-sided network. Unsponsored network: A platform-mediated network that lacks a clear sponsor, evolving instead through the collective efforts of platform providers, network users, infrastructure vendors, regulators, and other parties Examples include the electric power grid and the Internet. User: See network user. Variety platform: Facilitates transactions between supply-side users, who offer goods or services that vary along some dimension relevant to demand-side users. The goods or services are often complements that must be consumed in tandem with a platform good possessed by demand-side users. Examples include DVD, the World Wide Web, and credit cards. One of four types in a taxonomy of based on platforms’ principal function. Willingness to pay: In the context of platform-mediated networks, the maximum amount that a user would pay to access a network. Winner take all: In a winner-take-all networked market, a single platform serves all users. 21 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts Appendix A: Delimitations Our focus on platform-mediated business networks that exhibit network effects delimits the scope of the companies we study in four ways. Supply Networks. The term “network” is often applied to a company’s full set of business partners, e.g., its suppliers and distributors. The focal firm’s management of this “ecosystem” may be crucial to competitive advantage.27 However, a supply network is defined as a platform-mediated network only when network effects are evident. Some business partnerships meet this test, but many do not. Consider, for instance, Toyota. Consistent with the ecosystem metaphor, Toyota has a symbiotic relationship with many suppliers; some serve as Toyota’s sole source and/or have Toyota as their only customer. However, interdependence in this case does not entail a cross-side network effect.28 Specifically, consumers will not find Toyota’s platform more attractive if Toyota attracts more suppliers. Consequently, Toyota’s ecosystem is not a platform-mediated network, even though Toyota’s product family is designed around a platform and the firm has a huge supplier network. Physical Networks. The term “network” is also applied to firms that rely on physical infrastructure with a networked architecture (e.g., gas pipelines, electricity grids, railroads, and cable television systems). However, some businesses that employ physical networks do not exhibit meaningful network effects, for example, natural gas utilities. Gas utilities enjoy powerful supply- related scale economies related to the density of their pipeline networks. After a utility extends lines into a neighborhood—which entails a big upfront investment—the cost of hooking up additional customers is modest. With respect to demand-related economies of scale: A gas pipeline’s tree-and-branch architecture facilitates only one-way, point-to-multipoint distribution from the utility to its customers. The architecture does not allow point-to-point exchanges of gas between customers at the network’s endpoints. In other words, the utility cannot function like a connectivity platform. Unlike electric utilities (which also employ a tree-and-branch architecture), natural gas utilities are not part of variety platforms. The availability of 120-volt AC electric power in almost every U.S. household elicits the production of a vast array of electric appliances, which in turn encourages the nearly universal availability of the power grid. By contrast, there are only a handful of household applications for natural gas (e.g., furnaces, ovens, clothes dryers) so variety-based network effects are weaker. Business vs. Social Networks. In studying social networks, sociologists look at the structure and strength of dyadic interpersonal relationships (i.e., “ties”) between individuals (i.e., “nodes”) with respect to attributes such as frequency of interaction (who talks to whom?), power (who influences whom?), and affinity (who likes whom?).29 The inherent “point-to-point” structure of social networks means they share some properties of connectivity networks. In fact, some connectivity platforms—for example, instant messaging services—enjoy strong network effects because they are built around their users’ social networks. However, most social networks are not organized by profit-seeking firms, so they are not platform-mediated networks. Many businesses leverage word-of-mouth recommendations that propagate quickly through social networks. Word-of-mouth recommendations should be viewed as network effects only when they involve expectations of reciprocity. Recall that with network effects, a new user’s participation makes the network more valuable for all existing users. If I switch to oil heat based on my neighbor’s 22 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 advice (“We prefer oil over gas—it seems safer”), my neighbor does not value her oil service more highly, unless she expects that I might someday provide her useful advice (“Company X serviced our boiler; we liked their work”). For some products (e.g., failure-prone PCs), user-to-user advice that travels through social networks is important and strengthens same-side network effects. Traditional Retailers. Is a typical brick-and-mortar or Internet-based retailer, say, the consumer electronics chain Circuit City, a platform that mediates a two-sided network?30 We previously defined users in platform-mediated networks as independent actors and defined firms who create and maintain these platforms as intermediaries. These definitions imply that traditional retailers like Circuit City should not be considered platform-mediated networks. Intermediaries are “go-betweens” who serve the two principals to a transaction; they are not principals themselves. Most traditional retailers are principals: they take title to goods and absorb inventory risk, and they take responsibility for pricing and merchandising the goods. Vendors forfeit their independence by transferring title and have little or no control over how a retailer presents or prices their goods.31 However, vendors retain independence with respect to one crucial decision: whether to sell their goods to a retailer in the first place. In Circuit City’s market, most vendors seek wide distribution and sell their goods to almost any interested retailer. In other markets, vendors must balance gains from extra sales against incremental costs incurred in supporting a retail outlet (e.g., a larger direct sales force; increased warehouse complexity). Finally, some vendors may withhold products from certain retailers due to concerns about brand impact (e.g., as with luxury goods vis-à-vis discounters). When most of a retailer’s potential vendors face difficult decisions about whether to make their goods available, the retailer functions more like a platform mediating a two-sided network; it must mobilize users on both sides. By contrast, if most vendors will sell their goods without hesitation, then the mobilization challenge is trivial. For example, from its inception Amazon was able to source virtually all books in print through established wholesalers and jobbers. Notwithstanding the Amazon example, vendors’ decisions about whether to sell to a retailer tend to be more difficult when a market is young. For example, in deciding whether to supply online stores such as Apple’s iTunes, music and film companies confront concerns about increased piracy (due to uncertainty about digital rights management solutions) as well as negative reactions from established retail partners (e.g., Blockbuster Video). For these reasons, iTunes displays some attributes of a platform-mediated network. In recent years, some mature retailers have been changing their business models to incorporate more features of platform-mediated networks. For example, Wal-Mart has been shifting conventional retailing functions to its suppliers, asking them to absorb inventory risk and handle in-store merchandising. Wal-Mart brokers a more direct connection between its suppliers and shoppers, giving firms like Procter & Gamble greater control over how their goods are sold. These examples suggest a continuum, with the traditional merchant model at one end point and the two-sided platform model at the other. Many real-world retailers fall somewhere in between these pure types. 23 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts Appendix B: Multiple Equilibria, Tipping, and Expectations Management in Networked Markets In economics, an equilibrium is a resting point where no marketplace participant can realize superior value by changing strategy, assuming that all other participants also follow their optimal strategies. New markets with network effects may have two stable equilibria, measured in terms of user adoption rates: one at zero demand and another at the full saturation of the potential user base. Prospective users’ expectations regarding the ultimate size of the network determine which equilibrium will prevail. Uncertainty about how technology, business models, regulation, and demand will evolve is often high in new markets with network effects. For this reason, collective expectations regarding future network size are fluid—especially when markets are fragmented and participants cannot communicate beliefs or coordinate behavior. With fluid expectations, networked markets can be “tippy.” Once they start to move towards one equilibrium or the other, momentum can become strong. These dynamics are illustrated in Figure 4, which is based on work by Professor Nicholas Economides.32 Panel A of Figure 4 shows a family of demand curves for a product that exhibits network effects. Each curve corresponds to a different level of collective expectations about the network’s ultimate size, as perceived by prospective users. For example, curve D1, which is based on a small expected network size, is derived by asking each prospective user, “How much of the product would you buy if its price was $1, $2, $3, etc. AND you knew that the ultimate size of the network would be N1?” Curves D2 and D3 are derived the same way, increasing the network’s assumed ultimate size to N2 and N3, respectively. Such demand curves can be elicited through market research, for example, through a conjoint study. Panel A depicts a family of demand curves, but at any point in time, only one curve governs users’ behavior, namely, the curve that corresponds to users’ current collective expectations regarding future network size. If expectations were to change—as they often do—then a different curve would apply. In Panel A, demand shifts outward from D1 to D2 in response to a doubling in the expected network size from N1 to N2. D3 reflects the response to another doubling in expected network size, from N2 to N3. Note that demand shifts outward at a reduced rate with each doubling, reflecting a diminution of network effects. Note also that all the demand curves pivot at the same point on the horizontal axis, NS. When the product is free, everyone wants it, regardless of their initial expectations for network size. Panel B shows the “fulfilled expectations” demand curve for the new product, which traces a path through the family of “normal” downward sloping demand curves depicted in Panel A. The fulfilled expectations curve intersects each normal demand curve at the price that would exactly fulfill demand, given users’ expectations for future network size relevant to that normal demand curve. Hence, for D2, the ultimate network size is expected to be N2, and users will actually purchase N2 units only when a price of P2 is offered. If suppliers offered a lower price, purchases would exceed N2, and expectations would NOT be fulfilled. 24 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 The fulfilled expectations demand curve has an inverted “U” shape because initial gains in network size are extremely valuable to potential users. However, as the expected network size approaches the saturation point, NS, network scale gains are less valuable to users. The fulfilled expectations curve slopes downward once traditional price elasticity effects dominate network effects. Panel C assumes that “perfect competition” prevails. That is, there are many suppliers; none are large enough to impact the market price through their production decisions; and low entry barriers guarantee that the market price (P) equals suppliers’ marginal cost (MC). If P ever exceeded MC, then profits would stimulate additional production by incumbent suppliers and new entrants. Competition would drive P back down to MC. P = MC thus represents the industry’s supply curve, which is horizontal under perfect competition. Figure 4: Multiple Equilibria in Networked Markets Panel A Panel B Panel C WTP WTP WTP D3 Fulfilled Expectations D2 Fulfilled Expectations Demand Curve D3 Demand Curve D2 P2 P3 P Marginal Cost P1 D1 D1 N1 N2 N3 Ns N1 N2 N3 N0 N1 N4 Network Size Network Size Network Size Source: Casewriter. The market has three equilibria: N0, N1, and N4. At N0, demand and supply both equal zero. The supply curve intersects the fulfilled expectations demand curve at N1 and N4. Formally, N1 is an equilibrium, but it is not a stable resting point. Anything that shifts expectations—remember that they are fluid—just a little below or little above N1 will set off a chain reaction that causes the market to tip to N0 or N4, respectively. To understand why, imagine that users’ collective expectations regarding future network size were just a tiny bit greater than N1—say, they equaled N1.1. The relevant normal demand curve would not be D1, but rather D1.1 (not depicted in Panel B), which is just slightly above D1. However, an expected network size of N1.1 cannot be fulfilled, because price exceeds marginal cost at the point where the fulfilled expectations curve intersects D1.1. If P > MC, incumbent producers and entrants, responding to profit opportunities, would increase supply to the point where D1.1 intersected P = MC. At this lower price, users would purchase more than N1.1. Consequently, the actual installed base would exceed N1.1, which was the original expectation for network size. We 25 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024. 807-049 Module Note—Platform-Mediated Networks: Definitions and Core Concepts can assume that potential users would observe this and recalibrate their expectations upward, say, to N1.2. Accordingly, their demand curve would shift outward to D1.2. The fulfilled expectations curve intersects the new demand curve, D1.2, at an even higher price than with D1.1. To fulfill user’s expectations, suppliers would again have to offer a price in excess of marginal cost. And again, this would not be sustainable: production volumes would increase in response to profit opportunities, driving prices back down to P = MC. Moving down D1.2 to the point where it intersects P = MC, users would boost their purchases beyond N1.2. So once again, they would need to recalibrate their expectations, which would shift their demand curve further outward. This dynamic would continue until the market “tipped” to N4. Sustaining an installed base larger than N4 would not be possible, because a market clearing price lower than marginal cost is not sustainable. Suppliers would exit until the market price equaled marginal cost and N4 was purchased. Moving in the other direction, if users’ initial expectations for future network size were just slightly below N1, say, at N0.9, then suppliers would exit the market until supply and demand equated on normal demand curve D0.9. At this point, however, actual purchases would fall short of N0.9, so users would recalibrate their expectations downward. Extending this logic, suppliers would continue to exit until demand and supply tipped to N0. This analysis shows that network markets are “tippy”: they can stall, or they can grow explosively. Expectations regarding the ultimate network size are enormously important because they determine which equilibrium will prevail. The analysis above assumes a fragmented structure on both sides of the market. An atomistic base of users cannot communicate expectations or coordinate behavior. Likewise, on the supplier side, the platform is shared and unsponsored. Suppliers cannot profitably internalize network externalities, nor is any firm large enough to influence marketplace expectations. Tipping dynamics can still prevail in less fragmented markets, so expectations remain important. However, when platforms are proprietary and/or sponsored, providers have a greater ability to influence marketplace expectations and often will invest heavily to do so. 26 This document is authorized for use only in Timothy Derdenger's 45-872 M&O M3 2024 at Carnegie Mellon University from Jan 2024 to Jul 2024.  Module Note—Platform-Mediated Networks: Definitions and Core Concepts 807-049 Appendix C: Calculating the Magnitude of Network Effects for Spreadsheet Software In a market that exhibits network effects, a platform provider must consider the present and future benefits of building its user base when making current period decisions about pricing and marketing spending. Absent network effects, a company maximizes current period profitability by setting its price and marketing spending at levels that equate current period marginal revenue and marginal cost. With network effects, however, optimization is less straightforward. Acquiring an additional user—call her “X”—in the current period yields revenue directly from X during that period, and also boosts—very slightly—the willingness to pay of all other users, who value a larger network. Other users’ WTP remains higher in future periods, until “X” exits the network. This increase should be factored into current period pricing and marketing spending decisions. These dynamics can be illustrated using a stylized example of a fictional company—let’s call it “Blossom”—selling spreadsheet software in 1989 (before most office productivity software was sold in bundled suites). Spreadsheet software is subject to a connectivity network effect: users value the ability to exchange files and require compatible software to do so. Estimates of this network effect, calculated by Professors Erik Brynjolfsson and Chris Kemerer, are employed below.33 Assume that Blossom’s product price equaled $360 and, excluding customer acquisition costs, its variable expenses for supplying an additional customer (for order processing, disk production, technical support, etc.) were $72, or 20% of revenue. Blossom’s software was not compatible with that of rival spreadsheet vendors; their users could not exchange files with Blo