From Playground to Battleground

So what exactly is the software industry? Computer science may have taught us about the underlying evolutionary science, and maybe even some basic principles of software design, but it was silent on the com­mercial end.

Because software evolves, developers must constantly change their focus to help each generation’s users achieve increasingly sophisticated tasks. The aspects of software, of software engineering, and of software management that appear central to the field’s definition change every few years. At any given point in time, though, three key questions define the software industry: What types of software comprise the contemporary commercial industry? Which of the industry’s characteristics drive con­sumer preferences? and How can firms profit from developing software? If these questions sound familiar, they should. They echo the questions in the discussion of Lyle Bowlin’s drive to unseat Amazon.com: Of course, back then we asked them in their generic form: What’s your product? What makes your product special? How are you going to use that “special quality” to generate profits? Here we see them in their software-specific incarnation. But their underlying truth remains untouched. If you want to make money, you need both a product respon­sive to consumer needs and a business plan capable of generating profits. Software companies must rethink all three questions every time that the translation frontier evolves to define a new generation of software.

The two broad categories of commercial software are platforms and applications. Platform programs sit at the translation frontier separating the human environment from the computing environment. Human users communicate with the platform directly; the platform begins the trans­lation downward to voltage levels. When the machine responds to trigger the upward translation chain, the platform conveys the response back to the user. The best-known examples of platforms are operating systems, such as DOS, Windows, OS/2, Unix, and Linux. Sometimes, though, users need more help translating their concerns down to voltage-level queries. They enlist the aid of application programs, whose specific capa­bilities allow them to function as modular additions sitting atop the platform. Popular types of application programs include word proces­sors, spreadsheets, and games.

As software evolves, each generation incorporates new technological innovations to shift the translation frontier further away from voltage and closer to English. Much as this evolution shifts the balance between hardware and software, it also shifts the balance between platforms and applications. Each technological generation incorporates more tasks that had been required to translate down from English into the chain growing upward from voltage. These shifts migrate tasks from applications to platforms to hardware and impel software development toward increas­ingly natural input languages and interfaces. As a result, the boundaries between platforms and applications shift with each successive generation of technology. These boundaries have changed multiple times within the brief history of the software industry. Commercial opportunities follow that technological lead; again, technology creates opportunities and economic incentives indicate which opportunities businesses choose to pursue.

Under any generation of the technology, successful communication between the platform and the applications is critical; without this last remaining translation, the entire system is useless.

Despite the consistent (and ongoing) changes that the software indus­try encounters as technology advances, some basic things never seem to change. Almost all computers have always housed exactly one platform but many applications. And virtually all users have always looked for computer systems that can accomplish a broad range of tasks. Users have thus tended to select a platform first, and to accept the de facto restric­tion to applications that can communicate with the platform they select.

That user decision shifts our attention from computer science to network economics. The choice of platforms defines both the virtual net­works of the software world and the second great entry barrier of the software market. The “applications barrier to entry” is a two-sided network effect. It arises because rational applications developers write programs that run on the largest platform network, thereby making it even more attractive to users. Users, in turn, purchase the platform that hosts the largest collection of interesting applications, that choice mini­mizes the de facto restrictions that they need accept. The two trends rein­force each other, making the largest platform subject to strong network effects. As Microsoft and the dot-coms all learned, this barrier works only when developers and/or consumers commit to network membership by sinking switching costs. The Internet barrier to entry proved spuri­ous; Microsoft’s applications barrier to entry is quite real.

Much as the unprofitability barrier complicated the software indus­try’s birth, this network barrier threatens its continued health. One of the great questions hovering over the information sector is how to nav­igate around this barrier. Our best antitrust experts tried to answer it and failed.

But even if they do, that story would lie in the future. We’re still looking at the eternal: the invariant forces that shape the software indus­try as it evolves from one generation of technology to the next. The crit­ical lessons here are that network barriers to entry are ubiquitous and that the hurdles they impose become higher as the products become more complex and the market nears saturation. What’s more, these barriers are natural consequences of rational behavior; independent developers and individual consumers make only the decisions that maximize their return—at least in the short run. And while consumer decisions drive developer decisions and vice versa, the platform developer can sit back and watch the value of its platform grow exponentially through the inde­pendent unpaid efforts of developers and consumers.

Well, perhaps “unpaid” is a bit extreme. After all, given how much a platform owner can gain from building a big network, a little signing bonus might be in order. The applications barrier to entry creates some interesting incentives. On the one hand, platform development is tough work. Software companies that develop powerful platforms want—and deserve—compensation for their efforts. Logically, a good platform should be expensive. As a platform becomes more expensive, though, fewer users will buy it. Fewer users buying it reduces the applications it will attract. And the fewer applications it attracts, the less valuable the network becomes. Thus, paradoxically, the more expensive the platform, the less valuable the network. But the flip side is also true.

This second model lies at the heart of the “free software” or “open source” movement, in many ways the hobbyists’ response to Gates’s open letter. Gates had posited that in the absence of profitability, no one would write good software. After all, why would anyone invest the time and effort necessary to develop software if they couldn’t sell it? The hobby­ists behind the open source movement had an answer: pride of author­ship. But their other answers might be more compelling to those who don’t think like artists: first, a huge talent pool of programmers sharing their abilities freely should produce some pretty powerful software; second, good software creates service-sector opportunities that develop­ers are best positioned to win. Now in all honesty, plenty of people find even these rationales to be wanting, including those like Gates himself who consider this model of software development to be the antithesis of capitalism. But some good capitalists, including those running IBM and Sun, disagree. They contend that open-source development is an impor­tant part of a competitive software industry.

From the perspective of entry barriers, though, open source promises an alternative revenue stream to platform developers who choose to grow their networks by circulating their software free of charge. It thus has the potential to resolve the ultimate paradox of network businesses: the inverse relationship between platform price and network value. Shapiro and Varian outlined this resolution as a unique business strat­egy available only in network industries:³ Take a loss to build your network, then generate revenues from locked-in customers in your after­market.

That said, adventures in the aftermarket can be tricky—as Kodak learned during the post-Chicago counterrevolution. But most aftermar­ket strategies are not problematic. After all, consumers should under­stand that they will be locked into a network defined by whichever platform they choose. As long as the platform owner is honest about the kinds of aftermarket costs that consumers are likely to incur and about the kinds of opportunities that they’re likely to find to address those needs, the market cops need show no concern.⁴

Apple and Microsoft both dealt with this network paradox back in the 1980s. While they approached it in different ways, both companies realized that platform pricing was much harder than application pricing, precisely because of the relative sizes of the aftermarket. The weaker network barriers in applications markets allow software companies to operate like other companies in other industries, who try to set their prices to maximize profits through the optimal combination of unit prices and volume sales.

Strategic considerations highlight the differences between the incen­tives of platform developers and those of application developers. They also split the software industry into two arenas with distinct incentive patterns. The most likely source of revenue to a platform developer comes from the sale of network access and related support and services. The most likely source of revenue to an application developer comes from software sales. At the moment, Microsoft is kingpin of both worlds. It rules the platform world with Windows and reigns over an important part of the applications world with Office.⁵

Microsoft won the office-software battle despite a protracted govern­ment investigation that progressed through much of the early 1990s. But little awareness of Microsoft’s tactics in office applications markets worked its way into the public consciousness. After all, these actions occurred long before the information sector took up residence on the front page. By the time that the papers made big news of Microsoft’s attempt to use Windows to make the Internet its own and thereby to emerge as the undisputed kingpin of the information sector, the software playground in which the youthful Micro-Soft had merrily launched Altair BASIC had become a deadly battleground bloodied from Microsoft’s many victories.