Six Strategies for Overcoming the "Chicken and Egg" Problem
in Complement-Based Network Effects Markets
Chris Dixon and Jeff Rhodes
For Prof. Tom Eisenman
Managing Network Businesses
May 4, 2003
Establishing a successful new platform is the closest thing to nirvana in the high-tech business world. Propelled forward by complement-based network effects, successful new platforms enjoy a virtuous cycle of rising volume and increasing margins. In some cases only the government can bring the stratospheric profits down to earth.
But for every Microsoft there are countless failed attempts that crashed before ever taking off. It is extremely hard to coordinate all the interdependent players and products that are required to launch a successful new platform in a market subject to complement-based networks effects. Providers of the base product may have significantly different financial and strategic incentives than providers of complement products. But unless consumers have confidence in the ability of both the platform's base and complement producers to provide a strong value proposition that will endure over time, the new platform is doomed to failure.
The ability for companies to achieve the necessary coordination in order to overcome the so-called "chicken and egg" startup problem is especially important now due to the rapid proliferation of technology through nearly all areas of economic activity. Fortunately, the last few decades offer many examples of successes and failures from which to draw general lessons for new platform launches in the future. Here we draw upon recent history to offer six strategies that we have gleaned to help new platform launches overcome the chicken and egg problem.
Signal long-term commitment to platform success and competitive pricing
When Microsoft launched their marketing campaign for the Xbox game console, they prominently promoted the fact that they had committed $500 million to the campaign. Why would consumers care how much Microsoft was planning to spend on marketing? Because in complement-based network effects markets, expectations about the success of the platform in the future are critical to spurring consumer adoption in the present.
When deciding whether or not to adopt a new technology platform, consumers base their decisions not only on the current price and availability of complements but on their expectations for the future price and availability of complements. Consumers need to be convinced that the platform will survive and that complements will be reasonably priced now and in the future. Base product producers must find ways to provide these assurances to consumers in order to overcome the chicken and egg startup problem. Companies can accomplish this by using branding and marketing to signal commitment, employing dynamic pricing on current products to reduce consumer adoption costs, and using open standards to signal ongoing competitive pricing in the future. Using these tactics in combination plays an important role in overcoming the chicken and egg problem for base product producers.
The most direct way to influence consumer expectations about the viability of the platform is through branding and marketing. In the Xbox example above, Microsoft publicly committed a large marketing spend to signal commitment to the product. Microsoft hopes that this commitment will lead consumers to believe in the long-term viability of the platform, encouraging them to purchase the Xbox console now. In addition to a large marketing spend on a single product, preservation of a strong pre-existing corporate brand can also signal long-term commitment. Many industry observers, for example, attribute the success of the IBM PC over technologically superior rivals to IBM's prestigious brand. Likewise, many incumbents use their brand value to dissuade customers from choosing rival platforms. Microsoft is known for announcing "vaporware" products that are released either much later than announced or never at all. When prospective customers hear that Microsoft is planning a rival product, they are likely to be dissuaded from purchasing a competitive product for fear of being "orphaned."
In many cases, however, the customer cannot be convinced of the platform's viability through marketing alone. In this case, the platform producer can overcome lingering consumer uncertainty by subsidizing the cost of the platform base product in order to reduce the consumer's adoption risk. This use of dynamic pricing to reduce consumer adoption costs has been a frequent tactic in the video game industry. Sony, Microsoft, and Nintendo lose money on the sale of their game consoles but compensate with licensing fees revenues from third party software vendors. The effect is to dynamically modify the cost to the consumer according to the number of complements he consumes - a good proxy for the value he derives from the platform and complements.
In addition to getting the current prices right, platform producers need to signal to consumers that, if they commit to a platform now, they won't be exploited by over-priced complements in the future. The first and most obvious way to signal this commitment is by making a firm pricing commitment through explicit contracts or advertisements. A subtler, and more binding, way to do this is to encourage multiple sourcing for complements. For example, car buyers are assured that spare parts for their car will be reasonably priced because there are multiple sources for auto parts.
An even stronger variant of second sourcing, usually found in the information technology industry, is to "open" the platform. Publishing product interface specifications is the weakest form of openness because platform vendors retain the right to change the interfaces in the future, in which case complementary software vendors have to weigh the costs of modifying their software versus exiting the market. In addition, vendors like Microsoft have exploited their platform interfaces to give advantages to their own complementary applications and thereby extract higher rents from customers.
Releasing source code is a much stronger form of openness. Nokia has recently made a move to sell their Series 60 handset operating system to other handset manufacturers, including direct competitors who are (understandably) suspicious of Nokia's long-term motives. In order to allay these suspicions, Nokia has made the source code open to the public. Even if Nokia decided later to close a newer version of the source code, the complements would have be tied to the "open" version of the code, and Nokia would have to compete all over again against their very own open code to retain the compatibility of those complementary products. Of course, openness comes at the price of control: if the platform does indeed take off, Nokia will likely never make attractive profits from it precisely because they have committed to openness and allowed competitive second sources.
Signaling long-term viability and quality through marketing, pricing, and openness can be particularly important for companies that are fighting both the chicken and egg start-up problem as well as other competitors that are addressing a similar market opportunity. In these competitive situations, companies are often racing to acquire customers and the goals are two-fold: to achieve a critical mass of users and complements for your platform in order to generate the positive feedback cycles that create barriers to entry for future competitors, and also to inspire "FUD" (fear, uncertainty, doubt) in potential customers for your competitor's platform in order to prevent the competitor from reaching critical mass. Signaling commitment to winning the long-term equilibrium outcomes via the methods described in this section are an important tactic in competitive racing and can theoretically lower the user set at which companies can achieve critical mass and begin to enjoy positive feedback cycles. A critical mass user set is not a fixed market share, but instead varies based on expectations about the size of the equilibrium user set and complement availability, and the velocity of consumer adoption.1 The greater are consumers' expectations for the size of the platform's equilibrium user set and the more quickly that consumers expect the equilibrium user set to be achieved, the more likely consumers will be to adopt the platform. If companies are able to create high consumer expectations for equilibrium user set and adoption velocity via commitment and pricing tactics, then positive feedback loops with more customers and complementors will begin at smaller market shares. In effect, companies can win the race by moving the finish line closer-in.
Use backwards and vertical compatibility to leverage existing complements
Backward compatibility is often associated with software platforms such as operating systems. Every PC operating system Microsoft has released since the early days of DOS has been backwardly compatible. This has maintained their vast selection of complements and has therefore been an important reason why Microsoft has not encountered chicken and egg startup problems when launching new versions of its Windows platform. Interestingly, Microsoft even used this strategy as a startup, before they had any independent software vendors creating complements for their operating system: the first version of DOS was compatible with the dominant operating system at the time, CP/M, an operating system that Microsoft did not own. Companies can use insights from Microsoft's continued success when they are developing strategies to overcome chicken and egg problems. When launching a new technology platform, decreasing backwards and sideways friction is generally to your advantage. The benefit of gaining the incumbent's complements generally outweighs the converse benefits to the incumbent.
Although backward compatibility is associated with software products, it applies to many other products and even services as well. Consider the case of online bill presentment services.2 Services such as PayMyBills.com and Yahoo! Bill Pay allows customers the convenience of paying their monthly bills online for a low monthly fee. The chicken and egg problem these services initially confronted was that they could only get merchants to supply online bills to their system once they had a sufficient customer base. The backwards compatibility that these services devised was to offer a service where customers could redirect their paper bills to the bill presentment service provider who would scan them and put them online. This effectively provided backward compatibility with the old method of supplying bills on paper. The online bill presentment startups that offered this backwards compatibility provided enough initial value to attract a base of customers. Once sufficient customers were acquired, it was easy to convince merchants to support their new bill payment platform in order to save money on printing, shipping, etc.
Although this strategy may seem obvious in retrospect, it wasn't obvious at the time to many of the startups in the space. A prominent bill presentment company that did not adopt this backward compatibility strategy was TransPoint, a joint venture by Microsoft and FirstData. Not surprisingly, TransPoint never succeeded in attracting a critical mass of merchants and customers.3
Sideways, or horizontal, compatibility is another strategy new or lagging platforms can adopt to gain ground against a dominant competitor. Sideways compatibility is defined as making a platform compatible with a competing platform either through a change in the first platform or through an intermediate "bridging" technology. It differs from backward compatibility primarily in that it commits the platforms to future compatibility even with new releases of the rival platform. Perhaps the most interesting strategic example is Sun's creation of the Java programming language. The idea behind Java was to create a new programming language that would, because of its technological advantages, appeal to application producers, and would at the same time make programs written in the language easily portable across platforms. The desired effect was to reduce the dominance of Microsoft Windows over Sun's Unix variant, Solaris, by taking away their advantage in complementary products. This incredibly ambitious strategy posed a serious threat to Microsoft, leading them to adopt arguably unfair business practices by blocking the distribution of the Java runtime environment. Startups facing the chicken and egg problem can use bridging technologies like Java both to immediately acquire an initial base of complements and to assure consumers that they will have a steady flow of complements in the future.
It should be noted, though, that when considering the issue of compatibility, potential downsides should be taken into account as well. Besides the direct cost of developing and producing compatibility, ensuring compatibility often constrains the platform's performance. Sony has tried a variety of creative ways to make CD players convenient for use during athletics. In the end, however, the physical dimensions of CDs proved too frustrating, causing Sony to attempt to replace the CD format with Minidiscs.
In some cases, moreover, backward compatibility can completely backfire. Circuit city made their Divx movie player format backwardly compatible with DVDs in order to ensure that there would be enough complements for consumers who bought Divx machines. As a result, complement producers went for the widest market and decided to primarily create DVDs. This made DVD the de facto standard, ensuring Divx's demise.4
In summary, compatibility can be a powerful tool for overcoming the chicken and egg problem, but the benefits of compatibility need to be weighed against not only the direct costs of creating the compatibility, but also the performance costs and the risk that complement producers will target the "lowest common denominator." The ideal conditions for creating compatibility are those that offer consumers and complement producers the best of both worlds: a significantly more attractive new platform with all the benefits of the old or competing platform.
Exploit irregular network topologies
The Linux operating system has seen remarkable success in the server operating systems market, with its market share growing at 60% in 2002 to about 15%. At the same time, Linux has captured only miniscule market share among desktop operating systems.5 The reason for this discrepancy lies in the fact that the complement-based network effects on the desktop side are much more difficult to overcome than those on the server side. On the server side, the majority of demand for complementary software applications is highly concentrated and is easily satisfied by a handful of open-source and commercial offerings. That is, the server side complement-based network effects are highly concentrated around certain complements, or nodes, in the network. In contrast, the desktop complement-based network effects are more uniformly distributed across a broad array of complements, making it more difficult for Linux to overcome the chicken and egg problem and establish Linux as a viable platform for the desktop.
Real-life networks are often very different from the uniformly distributed networks pictured in the idealized mathematical representations found in economics textbooks. Metcalfe's law - the idea that the value of a network grows exponentially with its size -- applies only to these idealized networks. In contrast, recent academic work has found that many real-life networks are "scale-free," meaning that network connections tend to concentrate around select hubs.6 As the network topology deviates from the idealization, so do the utility maximization strategies of the complement producers and consumers. Rather than focusing on providing a platform for the entire network, companies should direct startup efforts around concentrated network hubs in order to gain necessary traction to overcome the chicken and egg problem.
For example, online dating sites would seem, at first glance, to be the ultimate winner-take-all market because of the strong liquidity-based network effects (a close cousin to complement-based network effects). A naïve application of Metcalfe's law would lead us to assume that a startup would have no chance to challenge a dominant competitor like Match.com. In fact, however, many dating sites that were started after the market leaders were already in dominant positions have done very well by focusing on special groups that primarily prefer to date only other members of those groups. Jdate.com, for example, was started well after Match.com had already established a critical mass of users but has nonetheless thrived by focusing on Jewish-only customers who prefer to date only other Jewish customers. Jdate.com has exploited the fact that the dating network is actually a set of sub-networks that are to varying, but lesser, degrees connected to the rest of the network. Other startups have caught on to this approach and now analysts are predicting a widespread segmentation in the dating web site market.
Even within a supercharged complement-based network effects market like the desktop PC world, self-contained sub-networks exist where startups and minor players can survive and profit. Apple computer has survived by focusing on two self-contained ecosystems - graphic design and education. Apple consumers in these niches are willing to trade broad compatibility for niche-specific features. This trade off is acceptable because there are distinct groups of consumers and independent software vendors in the Apple niche markets.
In most cases distinct groups of consumers and complement producers are found in market segmentation that is vertical, by industry or domain of application. In the software world, this usually means enterprise markets, which is one important reason that startup software vendors such as Siebel, BEA, and Peoplesoft have found much more success in the enterprise market than in the consumer market, where Microsoft dominates. Occasionally distinct sub-networks can be found in market segmentation by traditional demographic attributes. The video game console market has supported two successful video game systems for the past 4 generations of systems by segmenting the marketplace by age. Consumers under the age of 13 have their own community of interests served by vendors like Nintendo, while the 13-and-over segment has been served by Sega, Sony, and more recently, Microsoft. Contrast this to handheld game market, dominated by 13 year olds and younger, where only one system, Nintendo's Gameboy, has succeeded.
For new platforms the best strategy to approaching strong complement-based network effects markets is to execute what Geoffrey Moore calls the "bowling pin strategy:"7 find a niche where the chicken and egg problem is more easily overcome and then find ways to leverage the niche to the broader market. Linux, for example, started as a low cost, reliable platform for web servers, mail servers, DNS hubs and print servers where communication standards are open and therefore proprietary network effects are low. Now, as it tries to move into the desktop market, where Microsoft's proprietary network effects are very high, its strong position on the server side will have positive spillover effects that will make its attack much easier. As a result of Linux's server side popularity, for example, developers are increasingly moving to Linux, thereby lowering the cost of Linux application development.8 In addition, Linux has gained credibility in the enterprise market from important market influencers such as IBM and Oracle. This has given the Linux development team a chance to thoroughly test the system's security and reliability and develop a rich feature set. Whether this will be sufficient to break Microsoft's desktop dominance remains to be seen, but it will certainly make it much more likely than if this challenge were attempted by an untested, unknown operating system upstart. Thus the trick in the bowling pin strategy is to find a niche where network effects are more easily overcome and spillover effects are strong--attack the niche and then leverage it to the broader market.
Influence the firms that produce vital complements
Sony and Philips, the companies that oversaw the successful launch of the compact disc (CD) technology in the early 1980's, followed the CD launch with the introduction of the digital audiotape (DAT) in 1987. The DAT offered CD sound quality and, in a significant improvement over CD technology, it also offered the ability to record music.9 Despite these improvements, the DAT never gained significant consumer adoption and ended as an embarrassing failure for Sony and Philips. DAT failed because Sony and Philips failed to influence record companies, which produced the key complements for a consumer music system. Record companies were extremely concerned that the recording capabilities of DAT would lead to widespread piracy. Sony and Philips did not adequately address these concerns before the DAT launch. As a result, record companies produced little recorded music in the DAT format, making the technological advancements of the DAT platform significantly less valuable to consumers. Sony finally reached an anti-piracy agreement with record companies in 1992, but by that time consumer expectations for the DAT platform were dampened sufficiently to doom the platform.10
Complement-based network products depend on a complement in addition to the base product in order to create value for consumers. Therefore, as discussed earlier, consumers' decisions on whether or not to buy the base product are strongly influenced by their beliefs about the availability and quality of complementary products. In the DAT example above, consumers did not buy the base product, the DAT player, because they did not believe that the complement producers, record companies, would sell music in the DAT format. Successful companies overcome this chicken-and-egg problem by placing as much emphasis on marketing and influencing complement producers as end consumers. Influence can be achieved through many tactics, as discussed below, but the critical outcome is that complement makers have a strong willingness to produce products that enhance the value of the base product and moreover, that consumers are aware of this commitment. When companies succeed in marketing to and influencing complement producers, they are taking an important step in overcoming the chicken-and-egg start-up problem.
Although Sony and Philips were unsuccessful with the DAT launch, they were extremely successful with the launch of the CD five years earlier. Both the DAT and the CD represented significant advances for consumer audio systems at the time of their launches. However, when Sony and Philips launched the CD, they did a significantly better job of influencing complement producers. Most importantly, they addressed the record companies' primary concern by making CDs piracy resistant (or so it seemed at the time). In addition, Philips was able to influence Polygram, a major record label, to release music in the CD format because Philips owned a 50 percent stake in Polygram.11 Finally, Sony and Philips provided the record companies with access to their manufacturing technology and plant in order to ensure an adequate supply of complementary products. As a result, nearly 650 music titles were available in CD format when the first CD players were released and the CD format has gone on to become the most popular format for consumer music consumption.12
When IBM launched the first personal computer (PC) in 1981, the microcomputer market was still an unproven entity. IBM had failed in two previous attempts to launch a microcomputer, the 5100 in 1975 and the 5110 in 1976.13 Part of the problem was that no "killer apps" had existed to convince consumers to adopt these previous platforms. Based on these experiences, IBM decided to encourage independent software vendors (ISVs) to write applications for the PC that were compatible with the PC's DOS operating system. In order to influence complementors and overcome the chicken-and-egg problem with the new PC platform, IBM commissioned software from ISVs and also engaged in various other forms of ISV collaboration and subsidization.14 This program resulted in several complementary applications, such as Lotus 1-2-3 and the WordStar word-processor, that drove consumer adoption of the IBM PC platform and made the launch an instant success.
Microsoft is well known for recognizing the importance of complement providers and then placing great emphasis on influencing these providers. Microsoft is currently in the process of launching its .NET web services platform and is actively using the strategy of influencing complementors. The .NET platform includes several product SKUs, but its base products can be thought of as server operating systems (OS) and middleware and the complements as applications that run on these OS and middleware layers. In order to market to complement producers, the first .NET product that Microsoft brought to market was Visual Studio .NET, a tool for application developers. Microsoft generates negligible revenues from sales of developer tools, but it invests heavily in these products because it knows that influencing these complementors will be critical to the successful launch of the .NET platform. Microsoft has also invested heavily in supporting the community of complementors through funding for several initiatives that support ISV developers, including online newsgroups, the Microsoft developers network online support, and outreach by senior Microsoft executives, including Bill Gates and Steve Ballmer. When asked how he would measure the success of the .NET platform launch, one senior Microsoft executive responded, "The only thing that matters is if we have won the hearts and minds of the [ISV] developer community."15
Provide a standalone value for the base product
Philips introduced the videodisc player (VDP) in 1979 as a competitor to the VCR. VDPs had slightly better picture quality than VCRs and had potentially lower hardware and software costs, owing to a simpler manufacturing process. However, the VCR had a 3-4 year head start on the VDP and had already developed an installed base of over one million units.16 Overcoming the VCR's head start in a market subject to complement-based network effects would have been extremely difficult for the VDP. The VDP lost any chance to compete because it did not provide users with the ability to record television programming and watch the programming at a later time. This "time-shifting" of television programming was a standalone value that enabled the VCR to overcome the chicken and egg startup problem when it was launched, as discussed below. Without this standalone value, the VDP was forced to attempt to overcome an insurmountable chicken and egg problem startup problem.
Companies can overcome the chicken and egg startup problem by providing a standalone value for their base product - that is, a value proposition for consumers to purchase the base product independent of any new complementary products that might be introduced. Companies that are able to provide this standalone value for their base products are able to cultivate a user set large enough to make the chicken and egg startup problem disappear or weaken substantially. The essence of the chicken and egg problem is that users do not want to purchase the base product unless they know that complements will be available to make the base product valuable. But if the base product is valuable in its own right, then consumers will make their purchase decisions based primarily on the value proposition of the base product, with the potential for future complementary products as "icing on the cake." This is a vastly easier sell than marketing a blended value proposition that consists of both base product attributes as well as the likelihood for a large set of high quality complementary products at some future date. If the standalone value proposition for the base product is strong enough, a user set will purchase the base product in large enough numbers to entice complement makers begin to produce complement products without complicated inducements or value transfers from the base product producer.
Providing a stand-alone use is the strategy that VCR producers used to achieve a successful launch and avoid fighting the difficult chicken and egg startup problem. Unlike the VDP, the VCR offered users the ability to time-shift television programming. In fact, when the VCR was launched this was the only application available because the market for pre-recorded videocassettes had not yet developed. The standalone value for the VCR "time-shifting television programming" was sufficiently strong to induce over a million people to purchase the product in the first 3-4 years after its launch. This installed user base of the VCR as a base product was sufficient to entice entrepreneurs to develop a market for pre-recorded videocassettes as complementary products in the late 1970's.17 The complement-based network effect that resulted improved the value of the base product, increased sales velocity for the base and complementary products, and ensured that the VCR would be a common feature in most American homes.
Integrate vertically into critical complements when supply is not certain
When IBM launched the PC in 1981, it famously eschewed vertical integration and relied on third party companies to provide most of they key hardware and software components, including the PC-DOS operating system from Microsoft and the 8086 microprocessor from Intel. In contrast, when Apple launched the Macintosh computer in 1984, it maintained tight vertical integration by designing and manufacturing most critical aspects of the hardware and software layers. Although at opposite ends of the vertical integration spectrum, both companies missed out on huge profit opportunities by choosing the wrong level of vertical integration in response to the chicken and egg startup problem presented by the complement-based network effects in the PC market.
To overcome the chicken and egg problem, companies must find a way to ensure an adequate supply, variety, and quality of complementary goods. They can attempt to achieve this by collaborating with third parties in several ways, as described in an earlier section. However, whenever the base product supplier is reliant upon third parties to produce complement goods that are critical to the system, a heightened level of uncertainty will remain regarding value creation and value capture for the base product supplier. By vertically integrating into the complement product as well as the base product, a company can attempt to ensure an adequate supply of both goods. However, vertical integration is certainly not always a dominant strategy for success, as witnessed by the Apple computer example. By remaining tightly integrated, Apple precluded market competition from providing the necessary variety of price-competitive complements and base products. Companies can use vertical integration in certain start-up situations to ensure a reliable supply of complements, as described below, but they must be wary of stifling market forces that would drive down costs and improve quality on critical complements.
Vertical integration can help companies internalize externalities that sometimes prevent coordination between producers of the base product and complement products. For example, HDTV benefits TV producers, who can sell a new generation of TV consoles, but is primarily just an expense to broadcasters, who are primarily shifting advertising sales from one medium to another little net gain. So even though HDTV is a net profit opportunity, the platform might not be a success unless broadcasters are able to capture some of the positive externalities that their broadcasts create for TV producers. As discussed earlier, Philips used vertical integration into Polygram to ensure the availability of CDs for the launch of the CD player. Philips stood to gain significantly from the increased sales of CD players that would be generated by an ample supply of CDs. By vertically integrating into Polygram, Philips was able to internalize this externality and thereby reduce adoption barriers for consumers.
Companies must be careful about when to pursue vertical integration as a strategy for overcoming the chicken and egg problem. It can provide a powerful way for platform providers to ensure coordination between different players in the system, but tight integration can also stifle competitive forces that would improve the platform's value proposition to consumers and thereby increase the odds for success. Companies are most likely to be successful with vertical integration when the platform technology is not yet good enough to serve the needs of mainstream consumers. For example, advanced handheld devices (e.g., PDA's, advanced cell phones) are still relatively new and have not yet fully satisfied mainstream consumer needs, which is why a tightly integrated company such as Research in Motion has been so successful with its Blackberry product. Eventually, the technology matures and standard interfaces get established between different layers in the technology stack. At that point the basis of competition usually shifts to price, time to market, and personalization. The beneficiaries then become the firms that supply components that help people compete on cost and that control interfaces. As this shift occurs, companies that attempt to remain tightly integrated will find themselves encountering competitive platforms that are lower priced and have more selection due to market forces.18
Companies attempting to launch a new platform in a complement-based network effects market face incredible coordination challenges in order to overcome the chicken and egg startup problem. However, the rewards from a successful launch are tremendous and quite often long-lived. There is no fixed solution to achieve the coordination necessary to overcome the startup problem, but addressing the six tactics discussed here is an important start for any company launching a new technology platform.
Dranove, David and Gandal, Neil. "Surviving a Standards War: Lessons Learned from the Life and Death of DIVX." Kellogg School of Management Essay, January 6, 2003, <http://spirit.tau.ac.il/public/gandal/divx.pdf>, accessed April 2003.
Grindley, Peter. Standards Strategy and Policy: Cases and Stories. Oxford, England: Oxford University Press, 1995.
Hidding, Gezinus J. and Williams, Jeffrey R. "Are there first-mover advantages in B2B eCommerce Technologies?" Proceedings of the Hawaii International Conference on Systems Sciences, January 6-9, 2003.
Moore, Geoffrey. Crossing the Chasm. New York, NY: Harper Collins.
Noe Thomas, and Parker, Geoffrey. "Winner Take All: Competition, Strategy, and the Structure of Returns in the Internet Economy." Tulane University, November 30, 2002.
Parker, Geoffrey G. and Van Alstyne, Marshall W. "Information Complements, Substitutes, and Strategic Product Design." Tulane University and University of Michigan, November 8, 2000.
Petreley, Nicholas. "Debunking the Linux-Windows Market-Share Myth." Linuxworld.com, March 14, 2003. <http://www.linuxworld.com/2003/0314.petreley.html>, accessed April 2003.
Rolfs, Jeffrey H. Bandwagon Effects in High-Technology Industries. Cambridge, MA: The MIT Press, 2001.
Shapiro, Carl and Varian, Hal R. Information Rules: A Strategic Guide to the Network Economy. Boston, MA: Harvard Business School Press, 1999.
Spolsky, Joel. "Strategy Letter II: Chicken and Egg Problems." JoelonSoftware.com, May 24, 2000. < http://www.joelonsoftware.com/articles/fog0000000054.html>, accessed April 2003.
Swoyer, Stephen. "IBM King of Shrinking Server Market Heap." Enterprise Systems, March 6, 2003, <http://www.esj.com/news/article.asp?EditorialsID=445>, accessed April 2003.