Here's an interesting paradox. The computer industry is widely acknowledged to be one of the most brutally competitive markets the world has ever seen. So why are most segments of the market dominated by a single supplier?
IBM's previous dominance of corporate data-processing is a classical example of this "winner takes all" phenomenon. IBM wasn't just bigger than its closest rivals, it was roughly twice as big as the rest of the industry put together. Although it's more than a decade since IBM lost control of the market, it is still much bigger than Intel and Microsoft put together. Today, a conglomeration of giants, including Intel, Microsoft, Cisco (communications equipment) and Hewlett-Packard (printers), dominate the markets IBM used to own, but there are hundreds of examples of programs, such as Adobe Photoshop and Autodesk's AutoCAD, that dominate their respective markets.
This is not a new phenomenon. Wang used to own the dedicated word-processing business, Cray owned supercomputing, and Novell's Netware dominated networking. Netscape had a monopoly market share of the browser market, while RealNetworks had almost all the streaming media market. Intuit owned the home accounting market with Quicken, and Creative Labs owned the PC audio business with the SoundBlaster.
Not every market is dominated by a single supplier, even in computing, but so many IT markets tip towards having one giant and a lot of also-rans, that there must be more to it than appears at first glance. Something must be going on, beyond technical merit, advertising and marketing, distribution, price and other traditional factors.
One obvious reason why some IT companies become dominant is a traditional one: economies of scale. The more you make, the cheaper it gets. This is unusually important in computing, because research and development, and the construction of manufacturing plants, can represent a significant slice of the price. If it costs $100m to get a product to market and you sell 100m of them, the development cost is $1 a copy. If you only sell a million, it's $100 a copy. Now ask yourself if you want to invest $2.5 billion in a chip plant to attack Intel.
Microsoft has exploited economies of scale to the hilt, typically launching products at half or a quarter of the price of rivals. When PC databases cost $400-500 (£250-300) each and dBase II owned the market, Microsoft came in with Access at a special offer price of $99. Vermeer sold its web development software, Front Page, for $695; Microsoft bought it and relaunched it at $149 (and threw in a $40 rebate coupon).
Academic research has shown that, in 1988-1995, prices in the applications markets where Microsoft competed fell by 60%, whereas in those where it did not compete, prices fell by just over 10%.
But it is certainly possible for products to dominate their market even if they are expensive or, indeed, if they are not. Adobe PhotoShop 7.0, for example, costs more than $600, but has no serious rivals in the professional market. In the GNU/Linux arena, the Apache Web server, OpenOffice suite, The Gimp graphics program and GNU C also have no serious competition, and they are free. There must be other issues besides price.
One increasingly popular explanation is the idea of "network effects", which was invoked in the US Justice Department's case against Microsoft. The idea is that something becomes more valuable as more people use it. The classical example is the telephone system. One phone is no use, and two phones are of limited use. If there are a million phones, however, then the number of possible conversations is vast. At some point on the way to 100m or 1bn phones, it becomes hard to live without one.
Fax machines, instant messaging and the internet have benefited from the same phenomenon. Clearly, network effects are important in computing, too, even if there is not literally a network in place. For example, if everybody is using the MP3 file format for music, then products that support MP3 files become more attractive. Their development then makes the MP3 format even more attractive, with the feedback creating a "virtuous" circle. Network effects thus help generate things that become an important part of "the whole product", including software and peripherals, magazines, training and support. (See Schofield on Saturday, January 25 2003.)
However, while network effects can support a format, they cannot prevent it from being displaced. Market dominance did not stop Wang word processors from being replaced by personal computers running WordStar for CP/M, which in turn gave way to WordPerfect for DOS and then Microsoft Word for Windows. All that is required is for the benefits of switching to be greater than the switching costs. Dominant products are therefore at greatest risk when users are switching platforms - the cost of switching hardware platforms makes it easy to switch software, too.
This phenomenon is visible in its purest form in the games console business, where successive platforms are rarely compatible with what went before. Thus we have seen a series of temporary monopolies, as Atari VCS consoles gave way to 8-bit Nintendo Entertainment Systems, followed by the triumph of the 16-bit Sega Genesis/MegaDrive, which was displaced by the 32-bit Sony PlayStation. With the PlayStation 2, Sony has been the first company to win two stages in a row.
Of course, users resist switching platforms, because the costs can be enormous. (You don't just have to buy a CD player, you have to buy all your favourite albums again.) Users prefer evolution to revolution. This is why Microsoft - which is keenly aware of this particular issue - is trying to execute its revolutions by evolutionary means, and thus win three platform battles in a row: DOS, Windows, .Net.
A third and much better way to dominate a market is by setting a standard. Users sensibly prefer agreed, open standards, even if they are negotiated by cartels in back rooms. Even Americans are now recognising that Europe did better by having a common standard for mobile phones (GSM) than the open competition that resulted in an expensive, incompatible mess.
Unfortunately, standards have a poor record in the computer industry. It's a standing joke: "of course we believe in standards: that's why we have so many of them". Standards committees are often dominated by representatives from the leading suppliers, who have a vested interest in getting their patented technologies and products included, even if it is to the disadvantage of users. And the sheer length and expense of the standards-making process means that users can't afford to wait for them. In other words, while users would prefer open standards, they vote with their wallets to create de facto market standards, as was the case with SoundBlaster compatibility.
But the truth is that computer companies don't want standards unless they own them. What they really want is "lock in", which avoids competition by preventing users from having a free choice. This phenomenon was much in evidence during the Unix wars at the end of the 1980s.
European and American governments were prepared to back Unix - defined as Posix - as an open standard in preference to the expensive mess created by proprietary competitors such as IBM's System/38 (AS/400), DEC's VMS, HP's MPE and so on. However, factional in-fighting and self-interest prevented these companies from either agreeing or implementing a compatible Unix platform with Unix suppliers, such as AT&T and Sun Microsystems, even though this was what users wanted.
The Unix wars left the door wide open to Microsoft Windows, not just on the desktop, but on the server, where its market share was zero. Microsoft enforced what the hardware manufacturers were desperate to avoid: a software standard that obliged them to compete on a more or less equal terms, and therefore cut their prices. IBM used to make a 90% gross margin on hardware, and even Apple made 55%. In the PC industry, Dell's gross margin is 17.6%.
As I explained last week, the great advantage of the computer industry's horizontal layering is that you get competition at every level: chips, operating systems, languages, applications and so on. This encourages innovation and drives prices down. The disadvantage is that it's difficult to get all the layers to work together in a coherent way when there is little or no formal standardisation. The market's compromise is evidently to choose "winners" at each level, to get most of the benefits of standardisation without sacrificing the benefits of innovation and competition. It is an uneasy and unsatisfactory compromise, but it works.
Borrowing from Winston Churchill on democracy, it's probably the worst way of managing an industry, except for all the others.
