07-104

Copyright © 2007 by Thomas Eisenmann, Geoffrey Parker, and Marshall Van Alstyne.

Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.

Platform Envelopment Thomas Eisenmann Geoffrey Parker Marshall Van Alstyne

Platform Envelopment

Thomas Eisenmann, Geoffrey Parker, Marshall Van Alstyne May 17, 2007

ABSTRACT

Due to network effects and switching costs, platform providers often become entrenched. To dislodge them, entrants generally must offer revolutionary products. We explore a second path to platform leadership change that does not rely on Schumpeterian creative destruction: platform envelopment. By leveraging common components and shared user relationships, one platform provider can move into another’s market, combining its own functionality with the target’s in a multi-platform bundle. Dominant firms otherwise sheltered from entry by standalone rivals may be vulnerable to an adjacent platform provider’s envelopment attack. We analyze conditions under which envelopment strategies are likely to succeed. Keywords: Platforms, network effects, bundling, two-sided markets, convergence Thomas Eisenmann Associate Professor Harvard Business School Rock Center 218 Boston, MA 02163 Phone 617-495-6980 Fax 617-495-3817 teisenmann@hbs.edu

Geoffrey Parker Associate Professor A. B. Freeman School of

Business Tulane University New Orleans, LA 70118 Phone 504-865-5472 Fax 504-865-6751 gparker@tulane.edu

Marshall Van Alstyne Associate Professor Boston University School

of Management 595 Commonwealth Ave. Boston, MA 02215 Phone 617-358-3571 Fax 617-353-5003 mva@bu.edu

1

Platform-mediated networks comprise a large and rapidly growing share of the global economy. They encompass network users—individuals or firms whose interactions are subject to network effects—along with intermediaries who organize platforms that provide infrastructure and rules to facilitate users’ interactions (Rochet & Tirole, 2003; Eisenmann et al., 2006; Evans et al., 2006). Examples are as diverse as video games, container shipping, credit cards, instant messaging, package delivery, web search, real estate brokerage, HMOs, shopping malls, DVDs, online dating services, travel reservation systems, stock exchanges, and barcodes.

Established platform providers can be difficult to displace. They are sheltered by network effects (Farrell & Saloner, 1985; Katz & Shapiro, 1985; Economides, 1996), switching costs (Klemperer, 1987), and endogenous sunk costs for R&D, infrastructure, and marketing (Sutton, 1991). To dislodge entrenched platform providers, standalone entrants—firms that compete in a single product market—generally must offer revolutionary products and services (Henderson & Clark, 1990; Bresnahan, 1999). For these reasons, Evans & Schmalensee (2001) observed that networked industries often evolve through sequential winner-take-all battles, with superior new platforms replacing old ones.

This paper explores a second path to platform leadership change that does not rely on Schumpeterian creative destruction: a phenomenon we call platform envelopment. Platform providers that serve different networked markets often employ similar components and have overlapping user bases. We define platform envelopment as entry by one platform provider into another’s market, combining its own functionality with the target’s in a multi-platform bundle that leverages common components and/or shared user relationships. Dominant firms that otherwise are sheltered from entry by standalone rivals due to strong network effects and high switching costs may be vulnerable to an adjacent platform provider’s envelopment attack.

For example, Microsoft launched Windows Media Player (WMP) in 1998, enveloping RealNetwork’s dominant streaming media platform. RealNetworks (Real) served a two-sided network (Rochet & Tirole, 2003; Parker & Van Alstyne, 2005): consumers downloaded its media player for free and content companies paid for Real’s server software. Like Real, Microsoft freely supplied WMP to consumers, but Microsoft bundled its streaming media server software at no additional cost as a standard feature of Windows NT server—a multipurpose operating system that also incorporated file, print, mail, and web server software. Users found Microsoft’s bundles appealing and Real rapidly lost market share.

We believe that envelopment is a widespread phenomenon and a powerful force shaping platform evolution. While we do not offer statistical evidence of the incidence or impact of envelopment in this paper, we note that the phenomenon occurs in 13 out of 30 case studies in an MBA course on platform-mediated networks (Eisenmann, 2007). Examples include cable television system operators and local telephone companies each offering “triple play” bundles comprised of TV, phone, and Internet access services; eBay’s acquisition of PayPal; NTT DoCoMo’s move into mobile phone-based payment services; and LinkedIn adding job listings to its professional networking website to compete with Monster.com. Our analysis of motivations for envelopment draws heavily on senior management interviews completed in the context of these case studies.

2

In this paper, we present a taxonomy of different types of envelopment, based principally on the relationship between the attacker’s platform and that of its target; specifically, whether the platforms are complements, substitutes, or functionally unrelated. We explain the economic mechanisms behind different types of envelopment and describe conditions under which each is likely to be viable. Depending on the type of envelopment attack, bundling may yield price discrimination gains, efficiency improvements, or strategic advantages, for example, foreclosure of the attacker’s core market.

Contributions. Our main contribution is describing a mechanism for platform leadership change that does not require breakthrough innovation. In doing so, we integrate recent research on product bundling (e.g., Whinston, 1990; Bakos & Brynjolfsson, 2000; Carlton & Waldman, 2002; Nalebuff, 2004) and platform evolution (e.g., Fine, 1998; Bresnahan & Greenstein, 1999; Gawer & Cusumano, 2002; Iansiti & Levin, 2004).

In particular, we extend the work of Fine (1998) and Bresnahan & Greenstein (1999) and show how envelopment can drive epochal change when technical leadership is divided across industry layers, for example, microprocessors and operating systems in personal computers (Grove, 1996). Bresnahan (1999) observed that a dominant player in one layer would often sponsor entry into an adjacent layer, supplanting or diminishing that layer’s leaders. We show that cross-layer entry frequently will take the form of platform envelopment.

Our research also sheds light on why the boundaries between platform-mediated networks often blur. In industries that produce, process, and distribute digital information, such boundary blurring is called convergence. For example, handheld devices now incorporate the functionality of mobile phones, PCs, media players, video game players, and even credit cards. Convergence is a pervasive phenomenon that has received limited attention from scholars (Greenstein & Khanna, 1997; Yoffie, 1997). We assert that convergence is almost always the product of platform envelopment strategies.

Organization of the Paper. The balance of this paper is organized into four sections. Section 1 defines platform-mediated networks, explains why dominant platform providers often become entrenched, and describes different ways in which platforms may be related. Section 2 describes how economic and strategic benefits from bundling depend on the nature of the relationship between platforms. Section 3 presents a taxonomy of different types of envelopment attack based on the relationship between platforms, provides examples, and posits conditions under which each attack type is most likely to succeed. Section 4 considers issues for future research and methods appropriate for exploring these issues.

Platform-Mediated Networks In this section we define platform-mediated networks and describe some of their

properties. Next, we explain why established platforms—the targets of envelopment attacks—may be difficult to displace. Finally, we describe different ways in which two platforms may be related. These relationships are relevant to our analysis because

3

envelopment, by definition, involves one platform provider adding another platform’s functionality to its own.

Properties of Platform-Mediated Networks A platform-mediated network is comprised of users whose interactions are subject to

network effects, along with one or more intermediaries who organize a platform that facilitates users’ interactions (Rochet & Tirole, 2003; Eisenmann et al., 2006; Evans et al., 2006). The platform encompasses the set of components and rules employed in common in most interactions between network users (Wheelright & Clark, 1992). Components include hardware, software, and service modules, along with an architecture that specifies how they fit together (Henderson & Clark, 1990). Rules encompass information visible to network participants that is used to coordinate their activities (Baldwin & Clark, 2000). In particular, rules include standards that ensure compatibility between different components; protocols that govern information exchange; policies that constrain network user behavior; and contracts that specify terms of trade and the rights and responsibilities of network participants.

In traditional industries, bilateral exchanges follow a linear path as vendors purchase inputs, transform them, and sell output. By contrast, exchanges in platform-mediated networks have a triangular structure (see Figure 1). Network users transact with each other and they simultaneously affiliate with platform providers. For example, video game networks have two distinct groups of users, that is, two “sides.” Developers on the network’s “supply side” offer games to consumers on the “demand side”—the first set of bilateral exchanges. Developers must also contract with the platform’s provider—Sony Playstation, Microsoft Xbox, or Nintendo Wii—for permission to publish games and for production support—the second set of exchanges. Finally, consumers must procure a console and associated services from the platform provider—the third set of exchanges.

Every platform-mediated network has a focal platform at its core, although other platforms may play subordinate roles in the network, as explained below. The focal platform may have a single provider (e.g., Microsoft’s Xbox, eBay, Miami Yellow Pages). Alternatively, multiple providers may offer competing but compatible versions of a shared platform (e.g., Citibank Visa vs. Bank of America Visa; Ubuntu Linux vs. Red Hat Linux). If users switch between rival providers of a shared platform, they do not forfeit platform-specific investments, because all providers follow the same rules.

Platform Entrenchment Platform-mediated markets—also referred to here as networked markets—are

comprised of sets of platform-mediated networks. For example, the console-based video game market includes the Xbox, Playstation, and Wii networks; the U.S. credit card market includes the Visa, Mastercard, and American Express networks. Two platforms are rivals in the same networked market if they employ different rules and if changing the price that users pay to affiliate with one platform influences the other’s transaction volumes.

4

Figure 1: Elements of a platform-mediated network

Networked markets are typically served by only a few rival platforms; in many cases, almost all users rely on a single platform (e.g., Microsoft’s Windows, Adobe’s PDF, eBay’s online auctions). The number of platforms serving a networked market tends to be small when network effects are strong, multi-homing costs are high, and user demand for differentiated platform functionality is limited (Arthur, 1989; Caillaud & Jullien, 2003; Ellison & Fudenberg, 2003; Noe & Parker, 2005).

When network effects are positive and strong, users will converge on fewer platforms; a sub-scale platform will have little appeal, unless it provides the only way to interact with certain transaction partners. Likewise, users are less likely to multi-home—that is, rely upon multiple platforms—when it is expensive to establish and maintain platform affiliations. Finally, fewer platforms will be viable if users have relatively homogeneous needs. By contrast, if various user segments have distinct preferences and no single platform can profitably satisfy all segments’ needs, then the market is more likely to be served by multiple rival platforms.

With few rivals, established platform providers enjoy market power. High returns would normally attract entrants, but incumbent platform providers are often well protected. Factors that restrict the number of platforms in the first place can make it difficult and expensive to develop a new platform. Confronted with these barriers, most standalone entrants can only succeed if they offer significant improvements in performance and if they invest heavily to shift users’ expectations and absorb switching costs.

Relationships Between Platforms From the perspective of a focal platform’s users (for convenience, platform “A”), a

second platform (“B”) must be related to the focal platform in one of four ways. First, the two platforms may be functionally unrelated. Second, the two platforms may be

5

substitutes. Third, platform B may a nested network user of platform A. Finally, platform B may be a nested component employed by platform A’s users. Below, we explain these relationships and provide examples. In the next section, we show how the relationships affect benefits from bundling two platforms and thus the likelihood of successful envelopment.

Unrelated Platforms. Two platforms may be functionally unrelated, that is, designed to serve fundamentally different purposes, as with mobile phones and handheld video game devices. However, functional independence does not rule out the possibility that two platforms have common components or shared user relationships. For example, a handheld gaming device and a mobile phone each require a display, battery, microprocessor, and keys for input. Likewise, many teenagers own both a handheld gaming device and a mobile phone.

Substitutes. Rival platforms are close substitutes when they offer very similar functionality. Close substitutes are outside our scope of inquiry because envelopment, by definition, bundles two platforms with different functionality.

Rival platforms are weak substitutes—and candidates for envelopment—when they serve the same basic purpose, but employ technologies suited for different types of transactions. For example, Monster.com and LinkedIn.com use different approaches in helping users find and fill jobs: searchable listings and social networking, respectively. These approaches offer different advantages: listings are valuable when parties wish to conduct a comprehensive search, whereas social networks provide a mutually-trusted third party’s assessment of fit. Depending on situational needs, some job seekers and recruiters may rely on both platforms (i.e., multi-home); others may use just one.

Nested Network Users. Platform-mediated networks are often organized hierarchically in layers (Grove, 1996; Farrell, Monroe & Saloner, 1998). Consequently, the same entity may simultaneously serve as a platform provider for one network and as a user in another. For example, eBay is the platform provider for its online auction network, and at the same time, one of the World Wide Web’s millions of network users. In turn, the World Wide Web and dozens of other “application layer” platforms (e.g., email, file transfer protocol, instant messaging) are nested network users with respect to the Internet, a “transport layer” platform.

Nested Components. Platform providers serve as network users’ primary point of contact with the platform, but they do not necessarily supply all of the platform’s components. Frequently, platform providers rely on third parties who sell components directly to network users. Sometimes, these components are platforms themselves. For example, eBay provides interfaces that allow auction sellers to accept several forms of payment, including credit cards and PayPal, an email-based payment service acquired by eBay in 2002. These payment services are all platforms that serve two-sided networks comprised of merchants and consumers.

Reciprocally and Unilaterally Specific Complements When platform B is a nested component of platform A, platform A’s users will view

the two platforms as complements. From the perspective of A’s users, the platforms will be more valuable when consumed together than separately, and each platform will exhibit

6

negative cross-price elasticity with respect to the other. Likewise, when platform B is a nested supply-side user of platform A (e.g., a specific video game vis-à-vis a console), then platform A’s demand-side users (e.g., consumers) will view the two platforms as complements.

When two platforms are complements, the strength of cross-price elasticity between them may or may not be symmetrical. In some cases, the two platforms will be reciprocally specific complements: most use of each platform will be coupled with the other platform’s use (see Figure 2). For example, Microsoft Word—a platform that facilitates file exchange between users—is also a nested network user of Microsoft’s Windows operating system. Since Word is installed on a large fraction of PCs, and since Windows is the dominant PC operating system, Word and Windows are reciprocally specific complements. Each should exhibit reasonably high negative cross-price elasticity with respect to the other.

More typically, when platform B is a nested network user of platform A, B will be a unilaterally specific complement of A: most use of B will be coupled with A’s use, but the reverse will not be true.1 For example, most users of Quicken’s personal financial management software rely on Windows; the rest rely on Macintosh. However, only a modest fraction of Windows-compatible PC owners are Quicken users. Increasing the price of a unilaterally specific application like Quicken would have almost no impact on Windows’ sales, but increasing Windows’ price would reduce Quicken’s sales.

Figure 2: Reciprocally and unilaterally specific complements

As with nested network user relationships, nested component relationships can be reciprocally specific. For example, PayPal is used by 78% of eBay buyers; the balance pay with credit cards or money orders. With respect to PayPal transactions, 70% are

1 Unilaterally specific complements are similar to “one-way essential complements” in Chen & Nalebuff (2006). However, in Chen & Nalebuff, A is essential for all uses of B, whereas with unilaterally specific complements, A is associated with most uses of B.

7

completed on eBay; other online retailers and person-to-person payments account for the rest. Hence, the two platforms are reciprocally specific complements.

More typically, the focal platform will be a unilaterally specific complement to its nested component. For example, almost all eBay users require shipping services, but eBay represents only a small share of total shipments by Federal Express or UPS. Note that this relationship reverses the pattern of dependence normally seen with nested network users. Focal platforms are often asymmetrically dependent on their nested components, whereas nested network users are often asymmetrically dependent on their focal platforms.

We emphasize the distinction between reciprocally and unilaterally specific complements, because the strength and direction of relational dependence affects the payoff from bundling, and hence prospects for successful envelopment.

Bundling Benefits Envelopment entails one platform provider adding another platform’s functionality to

its own, and then offering a multi-platform bundle. In this section, we review potential benefits from bundling, which include price discrimination gains, efficiency improvements, and strategic advantages (Nalebuff, 2003; Carlton & Waldman, 2005a). We describe how bundling benefits depend on the nature of the relationship between platforms and posit conditions under which various benefits are likely to influence the odds of successful platform envelopment.

Price Discrimination Gains Bundling reduces heterogeneity in consumers’ aggregate valuations for a set of items,

allowing a firm with market power to extract a larger share of available surplus than it would earn from selling the items separately (Schmalensee, 1984; McAfee et al., 1989). Bundling thus serves as an effective price discrimination device in situations where a firm cannot assess customers’ willingness to pay for items or must offer the same price to all customers.

Compared to selling items separately, bundling is more likely to increase profits when: 1) the marginal cost of items is low or zero, as with many information goods; and 2) the correlation of consumers’ valuations for individual items is negative or weakly positive (Bakos & Brynjolfsson, 1999). If the correlation of consumers’ valuations is perfectly positive, then bundling does not reduce the heterogeneity of aggregate valuations and therefore does not yield price discrimination gains; bundling and selling items separately generate equal profits.

If marginal costs are material and the correlation of users’ valuations is not perfectly positive, then willingness to pay for a pure bundle (i.e., an offer of A+B, but neither A nor B alone) may fall below the bundle’s cost for a significant fraction of potential customers—those with little interest in one or both items. In such situations, a mixed bundle (i.e., offering A+B, A alone, and B alone) or tying (i.e., offering A+B and A alone but not B alone) may be more profitable than selling items separately—provided that the fixed cost of creating and marketing a bundle is not too high (Adams & Yellen, 1976; McAfee et al., 1989; Matutes & Regibeau, 1992).

8

Figure 3 summarizes analysis that builds on Nalebuff (2004) to illustrate conditions under which a pure bundle is more profitable due to price discrimination gains than selling the items separately. For a monopolist selling two items with uncorrelated valuations, demand is assumed to be uniformly distributed as prices range from zero to maximum valuations of VA and VB, respectively. The items are also assumed to have the same ratio of marginal cost to maximum valuation, i.e., MCA/VA = MCB/VB (see Appendix for details on model assumptions and derivations to support Figure 3). The horizontal axis of Figure 3 reflects VA/VB, that is, the ratio of the items’ maximum valuations, arbitrarily designating the item with the lower maximum valuation as A. The vertical axis reflects (MCA + MCB)/(VA + VB), that is, the ratio of the sum of the items’ marginal costs to the sum of their maximum valuations. Figure 3’s curve shows combinations of the marginal cost ratio and the ratio of the item’s maximum valuations for which bundling and selling separately yield equal profit. Selling separately yields greater profit than bundling for points above the curve, and less profit for points below the curve.

Figure 3: Attributes favoring bundling versus separate sales for platforms with uncorrelated valuations

When A has a very low maximum valuation relative to B, bundling is only profitable

when the items also have very low marginal costs. For items with equal maximum valuation (i.e., VA/VB = 1), bundling is more profitable than selling separately when (MCA + MCB)/(VA + VB) < 0.139. Expressed as the ratio of marginal cost to the profit-maximizing prices for the items sold separately, this threshold equates to (MCA + MCB)/((VA + MCA)/2 + ((VB + MCB)/2) < 0.243. Most information goods and many other platforms have marginal costs below this threshold. However, the threshold level of marginal cost for profitable bundling declines with the strength of positive correlation of

9

users’ valuations for two platforms (Bakos & Brynjolfsson, 1999). As explained immediately below, the type of relationship between the platforms systematically influences the correlation of users’ valuations.

Unrelated Platforms. Potential users’ valuations of two functionally unrelated platforms would not normally be correlated, unless use of each platform is associated with common underlying demographic or behavioral traits such as income, age, or profession. With common underlying traits, correlation is likely to be positive but not perfect, so price discrimination gains should be available. For example, many car owners value both emergency roadside services and auto loan brokerage. By bundling these and other functionally unrelated services, the American Automobile Association exploits price discrimination opportunities.

Weak Substitutes. When items in a bundle are complements or functionally unrelated, a user’s valuation for the bundle equals the sum of that user’s valuation for the items purchased separately. However, additivity does not generally hold for two perfect substitutes (Bakos & Brynjolfsson, 2000). Having access to a second perfect substitute provides no additional utility in the case of subscription services or perishable items, but some value with depreciable durables or consumable items that can be held inventory.

With two weak substitutes, a typical user’s valuation for a bundle should exceed her standalone valuation for the single item she prefers most, assuming that the two platforms each provide some functionality not available from the other. However, in terms of Figure 3, the ratio VA/VB is likely to be low, because a fraction of platform A’s functionality will be duplicated by platform B. Furthermore, demand for each platform’s unique functionality will often be positively correlated. For example, film fans who rent many movies will value both the wide variety available through mail delivery services and the instant access offered by Internet download services. Due to the asymmetry and positive correlation of platform valuations, bundling weak substitutes will only yield increased profits through price discrimination gains when marginal costs are very low.

Complements. Since reciprocally specific complements are typically consumed in tandem, a user with strong demand for one platform should value the other highly. Due to the strong positive correlation of users’ valuations, bundling the platforms is not likely to yield increased profits through price discrimination, although other bundling benefits may be available, as explained below. For unilaterally specific complements, the correlation of valuations will be positive but weaker, so price discrimination gains may be forthcoming from bundling.

Following from this analysis, we have: Proposition 1: Assuming marginal costs are low, bundling two platforms is more likely to yield increased profits through price discrimination when:

a. Platforms are functionally independent, and their use is not associated with common underlying demographic or behavioral traits.

b. Platforms are complements, but their use is not reciprocally specific.

10

Efficiency Improvements Compared to selling separate items, bundling can improve efficiency through

economies of scope in initial marketing; economies of scope in production and ongoing operations; quality advantages through an integrated design; and avoidance of double marginalization of perfect complements (Davis et al., 2001; Evans & Salinger, 2005)

Economies of Scope in Initial Marketing. When they buy bundled products, users can reduce search costs. Equivalently, firms that bundle products should realize economies of scope in customer acquisition activities, because they can sell the bundle with a single message. Likewise, bundling should reduce distribution costs, compared to shipping and stocking items that are sold separately.

Since customer acquisition and distribution expenses often have a large variable component, efficiency improvements will reduce marginal costs and increase the set of platform pairs that are candidates for profitable bundling. In terms of Figure 3, MCA + MCB for an A + B bundle will be lower than MCA + MCB for A and B sold separately. The magnitude of economies of scope in initial marketing does not depend systematically on the type of relationship between two platforms. Bundling should yield such savings regardless of whether the platforms are functionally unrelated, weak substitutes, or complements.

Economies of Scope in Production and Ongoing Operations. Compared to two products sold separately, an integrated design can reduce production costs by leveraging shared components. For example, video game consoles and DVD players both incorporate optical disk readers and circuitry for outputting audio and video to a television.

When two platforms are functionally unrelated, as with video games and DVDs, economies of scope in production and ongoing operations may be forthcoming, but it is not possible to generalize about their incidence or magnitude. However, the likelihood of realizing operating cost synergies through bundling does vary systematically when pairs of platforms are weak substitutes or complements.

Weak substitutes overlap to some extent in functionality, and hence should offer economies of scope in production and ongoing operations. For example, Netflix bundles two platforms that offer fundamentally different approaches for renting movies: mail delivery and Internet download services. Compared to firms offering the platforms separately, Netflix can save costs by relying on a single unit to procure films and by building a single website to facilitate users’ title selections.

By their nature, complements do not overlap much in functionality. Consequently, bundling two platforms typically will not yield major savings from sharing components. For example, by enveloping PayPal, eBay enjoyed some economies of scope in data center and fraud detection operations, but most functions provided by auction and payment services platforms were unduplicated.

Based on this analysis, we posit: Proposition 2: Bundling weak substitutes is likely to yield increased profits through economies of scope in production and ongoing operations.

11

Quality Advantages. Integrating two platforms may improve product quality through tighter management and/or simplification of component interfaces (Ulrich, 1995; Schilling, 2000). Improvement opportunities are especially likely for reciprocally specific complements that are consumed as a system, as with Apple’s iPod and iTunes. Integrating weak substitutes may also yield quality improvement. For example, by bundling DVD-by-mail and Internet download services, Netflix can amass a more complete transaction profile for a given user, and thereby offer better rental recommendations.

By contrast, integrating functionally unrelated platforms is more likely to result in design compromises that reduce quality, relative to platforms sold separately. For example, due to design flaws, consumers rejected Nokia’s N-Gage, which combined a mobile phone and handheld video game player. To accommodate gaming keypads, the device had a half-moon shape that was awkward for phone use. The N-Gage also required removal of its battery to insert game cartridges.

This logic yields: Proposition 3: Integrating reciprocally specific complements and weak substitutes may yield quality improvements compared to platforms provided by separate firms. Avoidance of Double Marginalization. Profits can be improved by avoiding double

marginalization when bundling complements that: 1) must be consumed in fixed proportions; and 2) would otherwise be supplied by separate monopolists who cannot price discriminate. Reciprocally specific complements often meet these conditions. For example, operating system (OS) and application software is consumed in one-to-one proportions.

In this context, a bundle’s optimal price will be lower than the sum of the prices offered by separate monopolists, who each will ignore the externality imposed upon the other (Nalebuff, 2000). Compared to traditional industries, the double marginalization of complements is more salient in networked industries, where winner-take-all outcomes are common (e.g., see Casadesus-Masanel & Yoffie, 2006, for analysis of double marginalization in the context of brinksmanship between Intel and Microsoft). Hence:

Proposition 4: Bundling reciprocally specific complements otherwise supplied by separate platform providers with significant market power may yield increased profits through avoidance of double marginalization.

Strategic Advantages The Chicago School’s “one monopoly price theorem” (OMPT) held that a monopolist

cannot increase profits by bundling its primary product with a complement (Posner, 1976; Bork, 1978). Whinston (1990) demonstrated that OMPT is valid under restrictive conditions: 1) the complement (denoted “B” in the discussion that follows) is supplied in a perfectly competitive market; 2) the monopolist’s product (denoted “A”) is essential for all uses of the complement; 3) the products are used in fixed proportions; and 4) the complement is subject to constant returns to scale. Carlton & Waldman (2005b) added a

12

fifth condition: OMPT requires a one-period setting, but may not apply with repeated purchases or upgrades over multiple periods, especially when switching costs are salient.

Platform-mediated markets often violate these conditions, so we would expect to observe profitable bundling of complements by platform providers. Specifically, as explained above, complements to a focal platform are often platforms themselves—either nested network users or nested components—and thus leverage network effects. Network effects may engender increasing returns (violating condition #4) and lead to an oligopolistic market structure (violating condition #1). Also, platform affiliation typically entails a long-term relationship for network users (violating condition #5).

Extending Market Power. If OMPT requirements are relaxed, a monopolist may be able to profitably extend its market power into a complement market (Whinston, 1990; Farrell & Weiser, 2003). According to Whinston (2001), when a monopolist’s primary product is not essential for all uses of a complement whose suppliers have market power, the monopolist has an incentive to exclude rival complement suppliers, to capture rent from alternate uses. When the complement market is subject to increasing returns—consistent with market power for standalone suppliers—the monopolist may have the ability to capture rents through bundling by denying scale to rival producers of the complement.

Specifically, assume that A is not essential for all uses of B, but the monopolist creates a bundle that ensures that A can only be consumed with its version of B, through either technical or contractual means. As an example, consider a counterfactual in which eBay, prior to acquiring PayPal, had banned that email payment service and mandated that auction participants instead use Billpoint, an eBay-owned payment service. Whinston’s conditions would be satisfied: PayPal enjoyed market power and increasing returns due to network effects. Also, as noted above, eBay’s near-monopoly in online auctions was not essential for all uses of PayPal.

In this scenario, standalone suppliers of B can make no sales to consumers who value an A + B bundle. In an alternative scenario, if rival versions of B remain compatible with A, then a standalone supplier of B will only be able to charge an amount equal to the value of any quality difference perceived by customers between its version of B and the monopolist’s. In either scenario, the demand reduction for standalone suppliers is greater to the extent that valuations for A and B are positively correlated and the bundle has strong appeal. Likewise, in both scenarios, standalone suppliers of B may not earn enough revenue to cover their fixed costs; they may be forced to exit the market, or may never enter in the first place.2 Matching the monopolist’s bundle would require

2 Bundling does not always weaken standalone suppliers. Under certain conditions, it may serve to soften rivalry and boost profits for all market participants. Carbajo, de Meza & Seidmann (1990) presented a model in which: 1) monopolized A is not essential for all uses of B; 2) undifferentiated versions of B are offered by A’s monopolist and by a single standalone supplier; 3) the correlation of customers’ heterogeneous valuations for A and B is perfectly positive; and 4) firms must offer, under Bertrand competition, the same price to all customers. Absent bundling, price competition would drive profits for the undifferentiated B products to zero. However, both firms can profit from the B market if the monopolist targets high-valuation customers with a high-priced A + B bundle and cedes low-valuation B customers to the standalone supplier. The bundle differentiates the B products; low-valuation customers cannot justify purchasing the high-priced bundle, so they only have one choice when buying B. This allows the standalone supplier to raise its price above marginal cost.

13

standalone suppliers to incur the presumably prohibitive cost of entering the monopolist’s primary market. This suggests:

Proposition 5: Through envelopment, a platform provider can profitably extend its market power when its target supplies either: 1) a reciprocally specific complement; or 2) a unilaterally specific complement that is used mainly but not exclusively with the attacker’s platform. Foreclosure of the Attacker’s Market. A dominant firm may also be able to weaken

existing rivals or deter entrants in its own market by bundling a complement (Whinston, 1990; Church & Gandal, 1992 and 2000). Consider a situation in which a dominant firm and its rivals all sell versions of A typically used in tandem with a reciprocally specific B complement, which is sold by a single standalone supplier subject to increasing returns. Consistent with the scenario above, if the dominant firm offers an A + B bundle, it denies revenue to the standalone supplier of B and may force it to exit. In that event, the dominant firm’s existing and potential rivals in the A market will lack a source of supply for the crucial B complement. This implies:

Proposition 6: Through envelopment, a platform provider can profitably weaken or deter rivals in its own market when its target supplies a reciprocally specific complement.

Taxonomy of Envelopment Attacks The analysis above suggests that the viability of platform envelopment depends

strongly on the relationship between two platforms. Accordingly, our taxonomy of envelopment attacks focuses on platform dyads that are functionally unrelated, weak substitutes, and complements. For each type of platform envelopment, we offer examples (see Table 1) and discuss the economic mechanisms that help or hinder an attack, specifically, the extent to which bundling engenders price discrimination gains, efficiency improvements, and/or strategic advantages (see Table 2).

Conglomeration Attacks In a conglomeration attack, the enveloper’s platform is functionally unrelated to the

target’s. The attacker’s bundle is a conglomerate, in the same sense as a corporation comprised of largely unrelated businesses or a rock consisting of separate stones cemented together.

Price Discrimination Gains. If use of two functionally unrelated platforms is not associated with common underlying demographic or behavioral traits, then potential users’ valuations for the platforms should be uncorrelated and significant price discrimination benefits may be available, provided that marginal costs for the platforms are also low (Proposition 1a).

Efficiency Improvements. Compared to providing platforms through separate firms, conglomeration attacks offer economies of scope in initial marketing, and in some cases, significant reductions in production and/or ongoing operating costs.

For example, cable TV system operators and local telephone companies have launched conglomeration attacks on each other. Cable companies now offer phone

14

service and phone companies have added TV service. These functionally unrelated platforms meet different needs but have significant user overlap. About 85% of U.S. households subscribe to multi-channel TV service through cable or satellite providers, and 95% subscribe to local phone service through fixed-line, mobile, or Voice over Internet Protocol providers. Consequently, bundling cable and phone services offers significant economies of scope in initial marketing.

Table 1: Examples of Platform Envelopment by Attack Type

Attacker Target Comments

Conglomeration Attack Cable TV system operators

Local phone companies

Attacks have been reciprocated.

NTT DoCoMo mobile phone service

Traditional credit cards

DoCoMo allied with Sumitomo-Mitsui (Japan’s 2nd largest credit card company) to offer mobile phone-based payment services.

Apple iPhone Smart phones from Nokia, Motorola, etc.

iPhone bundles iPod functionality into a smart phone.

Intermodal Attack UPS Federal Express UPS added overnight delivery in 1982; Federal

Express reciprocated, acquiring a trucking carrier in 1998.

LinkedIn’s online professional network

Monster.com’s recruitment platform

LinkedIn has added job listings, and Monster has added social networking functionality.

Blockbuster Video Netflix’s DVD-by-mail platform

Blockbuster added a mail-delivery option in 2004.

Foreclosure Attack Microsoft’s Windows RealNetwork’s media

player Windows Media Player displaced RealPlayer, but Microsoft paid RealNetworks $760 million to settle an antitrust lawsuit.

eBay’s auction marketplace

PayPal’s email payment service

eBay acquired PayPal after eBay’s in-house service, Billpoint, failed to displace PayPal.

Microsoft Office Adobe Acrobat PDF writer software

Office 2007 adds a “save as PDF” option, the core function of Adobe’s $299 Acrobat writer software.

15

Table 2: Motivations by Attack Type

Conglomeration Intermodal Foreclosure

Relationship Between Platforms

• Functionally unrelated

• Weak substitutes serving same basic purpose, but using technologies suited for different transaction types

• Complements, either reciprocally or unilaterally specific

Potential for Price Discrimination Gains (Assuming low marginal costs)

• Strong potential when correlation of users’ valuations is zero (P1a)

• Moderate potential when correlation of users’ valuations is weakly positive due to common underlying demographic or behavioral traits (P1a)

• Limited potential due to asymmetry and positive correlation of weak substitutes consumed in a bundle

• Limited potential for reciprocally specific complements due to strong positive correlation of users’ valuations

• Moderate potential for unilaterally specific complements due to weaker positive correlation of valuations (P1b)

Potential for Efficiency Improvements (Beyond economies in initial marketing available for all types)

• Significant economies of scope in ongoing operations available for some platform pairs but not others

• Operating cost synergies due to shared components and suppliers (P2)

• Quality improvements through superior knowledge of customer preferences across transaction types (P3)

• Quality improvements through tighter management or simplification of component interfaces (P3)

• Opportunity to avoid double marginalization when monopolists supply essential complements (P4)

Potential for Strategic Advantages

• Limited • Opportunity to neutralize an emerging threat

• Opportunity for attacker to extend its market power into a complement market (P5)

• Opportunity for attacker to protect its core market by foreclosing rivals’ access to reciprocally specific complements (P6)

16

The platforms also share many common components, so envelopers can capture operational synergies. Both cable TV and phone companies are able to utilize existing copper lines to deliver additional services without duplicating the huge upfront investment required to wire households. Also, both parties are able to leverage existing customer service operations when offering new services, including field technicians, call centers, and billing systems.

Strategic Advantages. Compared to other types of envelopment, indirect attacks offer fewer strategic advantages. Unlike a foreclosure attack (described below), an indirect attack cannot strengthen the enveloper’s core platform by denying its rivals access to a crucial complement. Also, unlike an intermodal attack (also described below), an indirect attack cannot neutralize an emerging substitution threat. Since strategic advantages are less salient, indirect attackers must rely on either price discrimination gains or efficiency improvements.

Convergence. Conglomeration attacks are frequently the mechanism behind convergence, a pervasive phenomenon in industries that produce, process, and distribute digital information. Yoffie (1997, p. 2) defines convergence as “the unification of function—the coming together of previously distinct products.” Convergence has several drivers, most notably performance improvements for semiconductors and broadband communications networks. Examples of digital convergence include: 1) Apple’s iPhone, which integrates a music player, mobile phone, and Internet communications device; 2) the Xbox 360 and PlayStation 3 video game consoles, which each include high-definition DVD players and Internet browsing capabilities; and 3) “triple plays” offered by cable and phone companies, incorporating not only the TV and telephone services mentioned above, but also broadband Internet access. All these examples of convergence are conglomeration attacks, bundling platforms that are largely functionally unrelated.

Intermodal Attacks With an intermodal attack, the enveloper’s platform and its target’s are weak

substitutes. They serve the same basic purpose but satisfy different sets of user needs, because they rely on fundamentally different technologies—that is, on distinct modes for accessing relevant functionality.

Price Discrimination Gains. Intermodal attacks are unlikely to yield price discrimination benefits due to the asymmetry and positive correlation of users’ valuations for weak substitutes consumed in a bundle.

Efficiency Improvements. Compared to providing platforms through separate firms, intermodal attacks offer economies of scope in initial marketing and—consistent with Proposition 2—potentially significant reductions in production and ongoing operating costs. Likewise, integrating weak substitutes may yield quality improvements, compared to two platforms offered by different firms (Proposition 3).

For example, prior to 1982, Federal Express and UPS fulfilled the same broad mandate—package shipping—but served different user needs. Federal Express relied on planes configured in a hub-and-spoke network to offer rush shipments of small, high-value packages. Prior to launching its own overnight air delivery service in 1982—an intermodal attack on Federal Express—UPS was dependent on trucks, which are better

17

suited for larger and less time-sensitive shipments. Many shippers send both types of packages, so UPS was able to capture efficiency gains through its envelopment attack. For example, UPS could market both types of service through a single sales force and could offer a single point of contact for customer service. While UPS did not realize long-haul transportation cost savings by offering both shipping modes, intermodal operations leveraged costs incurred in sorting packages and local delivery.

Strategic Advantages. When a new platform serves the same basic purpose as an established platform but offers different functionality, the potential for competition between the two platforms may initially be unclear. Given uncertainty about the extent of substitutability, an intermodal attack may be construed as a “better safe than sorry” strategy.

Netflix appears to be following this approach by bundling Internet downloads with its DVD-by-mail service. Likewise, Monster.com’s move into online professional networking was motivated in part by an emerging competitive threat from LinkedIn and other startups. By contrast, UPS missed its opportunity to neutralize a potential new rival. Federal Express was 12 years old and well established by the time it faced UPS’s intermodal attack.

Foreclosure Attacks With a foreclosure attack, the target is either a nested network user or a nested

component of the attacker’s platform. By bundling the target’s platform with its own, the attacker forecloses access to its users.

Price Discrimination Gains. Price discrimination gains are likely to be limited for reciprocally specific complements, due to the strong positive correlation of potential users’ valuations. When a target’s platform is a unilaterally specific complement, the correlation of potential users’ valuations should be weaker and an enveloper may realize price discrimination benefits from bundling, consistent with Proposition 1b.

Efficiency Improvements. Foreclosure attacks offer economies of scope in initial marketing, but economies of scope in ongoing operations are likely to be more modest. With reciprocally specific complements, an integrated design may yield quality improvements, compared to relying on a standalone supplier for a crucial complement (Proposition 3). Also, when bundling reciprocally specific complements otherwise supplied by independent monopolists, a foreclosure attack may also accrue efficiency gains by avoiding double marginalization (Proposition 4).

Strategic Advantages. A foreclosure attack limits access to users of the enveloper’s core platform, thereby reducing the target’s revenue. In this manner, the enveloper may extend its market power into the complement market, consistent with Proposition 5. If the platforms are reciprocally specific, the enveloper may also increase entry barriers in its core market, consistent with Proposition 6. When the target’s platform declines in value due to network effects, the enveloper’s standalone rivals must rely on an inferior complement.

Microsoft’s attack on RealNetworks’ streaming media platform, described in the introduction, is an example of a foreclosure attack that weakened Microsoft’s operating

18

system rivals. The compatibility of Windows Media Player with other PC operating systems was intermittent; for example, new versions did not work with Apple’s Macintosh prior to 2000 or after 2003. As a result, Microsoft bolstered its core Windows operating system by enveloping a crucial complement. However, Microsoft’s streaming media strategy drew antitrust scrutiny.

Cross-Layer Competition. Foreclosure attacks like Microsoft’s can play a powerful role in the evolution of multi-layer industries. As platforms mature, their providers sometimes embrace modular technologies and cede responsibility for supplying certain complements to partners (Langlois & Robertson, 1992; Suarez & Utterback, 1995; Baldwin & Clark, 2000; Jacobides, 2005; Boudreau, 2006). Often, these complements are platforms themselves. Reduced integration may result in an industry comprised of multiple layers, each with a separate set of suppliers. In the personal computer industry, for example, the layers consist of semiconductor manufacturing, PC assembly, operating system provision, and application software, among others (Grove, 1996). The credit card and telecommunications industries have similarly layered structures (Fransman, 2002; Evans & Schmalensee, 2005).

Over time, dominant players typically emerge within layers that are subject to strong scale economies due to fixed costs and/or network effects. These powerful players will extract a greater share of industry rents and vie with the focal platform’s original provider for technical leadership (Bresnahan, 1999). Friction over divided technical leadership is exacerbated as new layers with new leaders emerge. With a modular architecture conducive to experimentation, technological change may yield new uses for a platform, for example, browsers or streaming media software vis-à-vis PC operating systems. The original platform’s sponsor may be slow to integrate the new functionality due to inherent uncertainty about technology and demand. By the time the original platform sponsor offers the functionality, network effects may have propelled the new platform’s pioneer to a dominant position, effectively creating another industry layer.

Fine (1998) and Bresnahan (1999) observed that in the computer industry, the dominant player in one layer often pursued strategies to supplant or diminish another layer’s leaders. In some cases, challenges took the form of proxy wars, with players sponsoring weaker allies in adjacent layers (e.g., Microsoft backing AMD). In other cases, challengers directly entered an adjacent layer. Direct entry usually took the form of a foreclosure attack, as with Microsoft’s respective envelopment attacks on RealNetworks’ streaming media software, Netscape’s browser, Adobe’s PDF standard, and TiVo’s digital video recorder platform. More recently, industry observers have speculated that Google might mount a cross-layer foreclosure attack on Microsoft Windows by offering a Linux-based operating system with tight linkages to Google’s applications. Likewise, Google has bundled its paid search platform with a payment service (Google “Checkout”) and a product listing service (Google “Base”) in a cross-layer foreclosure attack against eBay.

Issues for Future Research This section considers three sets of issues for future research on platform

envelopment. The first asks when envelopers should acquire an incumbent platform provider rather than build a new platform internally. The second set concerns defensive

19

options for firms targeted for envelopment. For example, when should they match the attacker’s bundle? The third set of issues examines organizational challenges posed by envelopment attacks, which require a high level of internal, cross-unit coordination. The paper concludes with a discussion of research methods appropriate for future work on platform envelopment.

Acquisition vs. Internal Development Future research should analyze tradeoffs confronting envelopers who can choose

between acquiring an incumbent platform provider and developing a platform de novo. By acquiring a going concern, an enveloper can speed its attack, especially when it lacks crucial skills and resources. Also, an acquisition can avoid the risk that mounting losses, coupled with pressure from investors to improve short-term profit performance, will lead to the premature shutdown of an in-house startup. However, compared to building from scratch, acquiring an existing platform is likely to result in substantial integration challenges that may result in compatibility problems in the near term.

Defensive Strategies Researchers also should study factors that favor different defensive strategies for

firms facing envelopment. Besides investing aggressively in innovation to stay a step ahead of potential attackers, options include: • Changing Business Models. Simply ceding the targeted platform and redeploying

into new markets may be the best option for some firms. RealNetworks pursued this approach in response to Windows Media Player, scaling back investment in its streaming media platform. RealNetworks leveraged existing relationships with consumers and music companies to launch Rhapsody, an online subscription music service.

• Matching the Bundle. Firms may be able to counter an envelopment attack by assembling a comparable bundle. The target can cross parry (Karnani & Wernerfelt, 1985) if it has the requisite skills and resources to enter the attacker’s core market, and if barriers are not insurmountable due to intellectual property protection, high switching costs, or other factors. These dynamics are evident in online recruitment: as noted above, Monster.com added networking features in response to an envelopment threat from LinkedIn, which subsequently added job listings. However, bundle-versus-bundle competition can be exceptionally fierce (Bakos & Brynjolfsson, 2000; Nalebuff, 2000), so researchers should consider conditions under which accommodating an attack is likely to be more profitable than matching the enveloper’s bundle.

• Mergers and Alliances. Instead of assembling bundles on their own, platform providers targeted for envelopment can merge or joint venture with their attackers’ rivals. Lotus, vulnerable to Microsoft’s Office Suite, pursued this approach in merging with IBM. Likewise, RealNetworks’ Rhapsody platform, which is vulnerable to envelopment by mobile phone and Internet service providers, was able to co-opt potential rivals by joint venturing with Cingular and Comcast.

20

• Antitrust Litigation. Antitrust law for networked industries is still under dispute (Evans, 2003; Farrell & Weiser, 2003; Carlton & Waldman, 2005a). Dominant platform providers that offer bundles risk being charged with illegal tying. Several targets for envelopment by Microsoft have received favorable antitrust rulings and/or settlement payments, including Novell, RealNetworks, Sun Microsystems, and Netscape.

Organizational Issues Launching an envelopment attack requires an unusually high level of cross-unit

coordination. Engineers must integrate two platforms’ functionality and marketers must formulate joint pricing and targeting strategies. Even if the technology can be developed internally, most companies find that achieving cross-unit cooperation is difficult, because managers will fight for autonomy and strive to advance their units’ interests (Garud & Kumaraswamy, 1995; Galunic & Eisenhardt, 2001).

Future research could focus on optimal organizational structures for encouraging cross-unit cooperation in the context of envelopment, in particular, the appropriate degree of structural separation between new and old units (Tushman & O’Reilly, 1996; Christensen, 1997) and the ideal level of centralization of shared functions. Researchers also could study the organizational systems, processes, and cultural values that successful envelopers use to promote cross-unit cooperation, including transfer-pricing arrangements and incentive plans.

Methodological Issues Future research on envelopment should rely on a range of methods. Case studies

based on field interviews and analysis of archival records could reveal motivations for envelopment beyond those discussed in this paper. Qualitative methods would also be useful in exploring approaches that firms use to address the organizational issues described above.

Econometric analysis could test hypotheses about factors that increase a firm’s propensity to pursue envelopment and conditions that improve the odds of success with different attack types. Likewise, additional analytical modeling could shed light on these issues. To analyze the dynamics discussed above, however, a closed-form model would need to reflect heterogeneity and correlation in potential users’ valuations, economies of scope in marketing and production, and multi-homing costs, among other factors. Ideally, models would evaluate not only pure bundling, but also tying and mixed bundling, and would do so in a two-sided context. Due to the unusually large number of variables this would require, researchers will likely find it difficult to develop tractable closed-form models that capture the full richness of the envelopment phenomenon. However, modeling specific aspects may be feasible. Likewise, simulation techniques may be well suited for dealing with the intrinsic complexity of platform envelopment strategies.

Conclusion In this paper, we introduced the concept of platform envelopment. By leveraging

common components and shared user relationships, one platform provider can envelop

21

another’s market, combining its own functionality with the target’s in a multi-platform bundle. We presented a taxonomy of envelopment attacks, based on the relationship between the attacker’s platform and its target’s, and discussed the economic and strategic motivations for each attack type.

Briefly, with a conglomeration attack the platforms are functionally unrelated, but may share common users and components. In that event, bundling can yield significant economies of scope. With an intermodal attack the platforms are weak substitutes, and the attacker may neutralize an emerging competitive threat. With a foreclosure attack the platforms are complements. If envelopment reduces the appeal of a standalone provider’s crucial complement, the attacker may strengthen its position vis-à-vis rivals in its core market.

Envelopment is a widespread phenomenon that plays an important role in platform evolution. It is an effective strategy for mounting cross-layer challenges in industries where different firms dominate hierarchically organized strata of complements. Likewise, the pervasive convergence of businesses that produce, process, and distribute digital information often takes the form of platform envelopment. Most significantly, envelopment provides a mechanism for platform leadership change that does not require breakthrough innovation or Schumpeterian creative destruction. Well-established incumbents that otherwise are sheltered from entry by standalone rivals due to strong network effects and high switching costs may be vulnerable to envelopment.

22

References Adams W, Yellen, J. 1976. Commodity bundling and the burden of monopoly. Quarterly Journal of

Economics 90: 475-498.

Arthur W B. 1989. Competing technologies, increasing returns, and lock-in by historical events. Economic Journal 99: 116-131.

Bakos Y, Brynjolfsson E. 1999. Bundling information goods: Pricing, profits and efficiency. Management Science 45:1613-1630.

Bakos Y, Brynjolfsson E. 2000. Bundling and competition on the Internet: aggregation strategies for information goods. Marketing Science 19: 63-82.

Baldwin C, Clark K. 2000. Design Rules, Vol. 1: The Power of Modularity. MIT Press: Cambridge, MA.

Bork R. 1978. The Antitrust Paradox. Basic Books: New York.

Boudreau K. 2006. Does opening a platform generate more innovation? MIT Sloan Research Paper 4611-06.

Bresnahan T. 1999. New modes of competition: implications for the future structure of the computer industry. In Competition, Innovation and the Microsoft Monopoly: Antitrust in the Digital Marketplace. Eisenach J, Lenard T (eds.). Kluwer: Boston.

Bresnahan T, Greenstein S. 1999. Technological competition and the structure of the computer industry. Journal of Industrial Economics 47: 1-40.

Caillaud B, Jullien B. 2003. Chicken & egg: Competition among intermediation service providers. Rand Journal of Economics 34(2): 309-328.

Carbajo J, de Meza D, Seidmann D. 1990. A strategic motivation for commodity bundling. Journal of Industrial Economics. 38: 283-298.

Carlton D, Waldman M. 2002. The strategic use of tying to preserve and create market power in evolving industries. Rand Journal of Economics 33(2): 194-220.

Carlton D, Waldman M. 2005a. Theories of tying and implications for antitrust. Johnson School Research Paper 24-06.

Carlton D, Waldman M. 2005b. Tying, upgrades, and switching costs in durable-goods markets. NBER Working Paper 11407.

Casadesus-Masanell R, Yoffie, D. 2006. Wintel: cooperation or conflict. Management Science forthcoming.

Chen M, Nalebuff B. 2006. One-way essential complements. Working paper available at SSRN.

Church J, Gandal N. 1992. Integration, complementary products and variety. Journal of Economics and Management Strategy 1: 651-675.

Church J, Gandal N. 2000. Systems competition, vertical merger, and foreclosure. Journal of Economics and Management Strategy 9: 25-51.

Christensen C. 1997. The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail. Harvard Business School Press: Boston.

Davis S, MacCrisken J, Murphy K. 2001. Economic perspectives on software design: PC operating systems and platforms. NBER Working Paper W8411.

Economides N. 1996. The economics of networks. International Journal of Industrial Economics 14: 673-699.

23

Eisenmann T. 2007. Managing Networked Businesses: course overview for educators. HBS note #807-104.

Eisenmann T, Parker G, Van Alstyne M. 2006. Strategies for two-sided markets. Harvard Business Review Oct. 2006.

Ellison G, Fudenberg D. 2003. Knife-edge or plateau: when do market models tip? Quarterly Journal of Economics 118: 1249-1278.

Evans D. 2003. The antitrust economics of multi-sided platform industries. Yale Journal on Regulation 20: 325-382.

Evans D, Salinger M. 2005. Why do firms bundle and tie? Evidence from competitive markets and implications for tying law. Yale Journal on Regulation 22: 37-89.

Evans D, Hagiu A, Schmalensee R. 2006. Invisible Engines: How Software Platforms Drive Innovation and Transform Industries. MIT Press: Cambridge, MA.

Evans D, Schmalensee R. 2001. Some economic aspects of antitrust analysis in dynamically competitive industries. NBER working paper W8268.

Evans D, Schmalensee R. 2005 (2nd edition). Paying with Plastic: The Digital Revolution in Buying and Borrowing. MIT Press: Cambridge, MA.

Farrell J, Monroe H, Saloner G. 1998. The vertical organization of industry: systems competition versus component competition. Journal of Economics and Management Strategy 7: 143-182.

Farrell J, Saloner G. 1985. Standardization, compatibility, and innovation. Rand Journal of Economics 16: 70-83.

Farrell J, Weiser P. 2003. Modularity, vertical integration, and open-access policies: towards a convergence of antitrust and regulation in the Internet age. Harvard Journal of Law & Technology 17: 86-118.

Fine C. 1998. Clockspeed: Winning Industry Control in the Age of Temporary Advantage. Perseus Publishing: New York.

Fransman M. 2002. Telecoms in the Internet Age. Oxford University Press: London.

Galunic C, Eisenhardt K. Architectural innovation and modular corporate forms. Academy of Management Journal 44: 1229-1249.

Garud R, Kumaraswamy A. 1995. Technological and organizational designs to achieve economies of substitution. Strategic Management Journal 17: 93-109.

Gawer A, Cusumano M. 2002. Platform Leadership: How Intel, Microsoft, and Cisco Drive Industry Innovation. Harvard Business School Press: Boston.

Greenstein S, Khanna T. 1997. What does industry convergence mean? In Competing in the Age of Digital Convergence. Yoffie D. (ed.). Harvard Business School Press: Boston.

Grove A. 1996. Only the Paranoid Survive: How to Exploit the Crisis Points that Challenge Every Company and Career. Doubleday: New York.

Henderson R, Clark K. 1990. Architectural innovation: the reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly 35: 9-30.

Iansiti M, Levin R. 2004. The Keystone Advantage: What the New Dynamics of Business Ecosystems Mean for Strategy, Innovation, and Sustainability. Harvard Business School Press: Boston.

Jacobides M. 2005. Industry change through vertical disintegration: how and why markets emerged in mortgage banking. Academy of Management Journal 48: 465-498.

Karnani A, Wernerfelt B. 1985. Multiple point competition. Strategic Management Journal 6: 87-96.

24

Katz M, Shapiro C. 1985. Network externalities, competition, and compatibility. American Economic Review 75: 424-440.

Klemperer P. 1987. Markets with consumer switching costs. Quarterly Journal of Economics 102: 375-394.

Langlois R, Robertson P. 1992. Networks and innovation in a modular system: lessons from the microcomputer and stereo component industries. Research Policy 21: 297-313.

Matutes C, Regibeau P. 1992. Compatibility and bundling of complementary goods in a duopoly. Journal of Industrial Economics 40: 37-54.

McAfee R, McMillan J, Whinston M. 1989. Multiproduct monopoly, commodity bundling, and correlation of values. Quarterly Journal of Economics 114: 371-384

Nalebuff B. 2000. Competing against bundles. In Incentives, Organization, and Public Economics. Hammond P, Myles G. (eds). Oxford University Press: London.

Nalebuff B. 2003. Bundling, tying and portfolio effects. DTI Economics Paper no. 1.

Nalebuff B. 2004. Bundling as an entry deterrent. Quarterly Journal of Economics 119: 159-187.

Noe T, Parker G. 2005. Winner take all: Competition, strategy, and the structure of returns in the Internet economy. Journal of Economics and Management Strategy 14: 141-164.

Parker G, Van Alstyne M. 2005. Two-sided network effects: a theory of information product design. Management Science 51: 1494-1504.

Papoulis A. 1991 (3rd edition). Probability, Random Variables, and Stochastic Processes. McGraw-Hill: Boston.

Posner R. 1976. Antitrust Law: An Economic Perspective. University of Chicago Press: Chicago.

Rochet J, Tirole J. 2003. Platform competition in two-sided markets. Journal of the European Economic Association, 1(4): 990-1029.

Schilling M. 2000. Towards a general modular systems theory and its application to inter-firm product modularity. Academy of Management Review 25: 312-334.

Schmalensee R. 1984. Gaussian demand and commodity bundling. Journal of Business 57: 211-230.

Suarez F, Utterback J. 1995. Dominant designs and the survival of firms. Strategic Management Journal 16: 415-430.

Sutton J. 1991. Sunk Costs and Market Structure: Price Competition, Advertising, and the Evolution of Concentration. MIT Press: Cambridge, MA.

Tushman M, O'Reilly C. 1996. Ambidextrous organizations: managing evolutionary and revolutionary change. California Management Review 38(4): 8-30.

Ulrich K. 1995. The role of product architecture in the manufacturing firm. Research Policy 24: 419-440.

Wheelright S, Clark K. 1992. Revolutionizing Product Development. Free Press: New York.

Whinston, M. 1990. Tying, foreclosure and exclusion. American Economic Review 80: 837-859.

Whinston M. 2001. Exclusivity and tying in U.S. v Microsoft: what we know, and don’t know. Journal of Economic Perspectives 15: 63-80.

Yoffie D. 1997. Introduction: CHESS and competing in the age of digital convergence. In Competing in the Age of Digital Convergence. Yoffie D. (ed.). Harvard Business School Press: Boston.

25

26

27