Understanding software as a commodity

The "commoditisation of software" was one of the panel topics at the AO2004 Innovation Summit at Stanford University. By the end of the panel discussion, I didn't have a better idea of what commoditisation meant in relation to the software industry. It seemed to be an addendum to the dialogue that began with Nicholas Carr's Harvard Business Review essay, "IT Doesn't Matter", which argued that IT is trending toward commoditisation, meaning that organisations can't gain much strategic advantage versus competitors in applying technology to business problems.

During the panel exchange, Jonathan Schwartz, president and COO of Sun, quipped that "a commodity is a product that you don't ship and your competition does". On a more serious note, he defined a commodity as a product or service for which there is a universal or perpetual demand.

Marten Mickos, chief executive of MySQL, and Zach Nelson, chief executive of NetSuite, argued from the customer perspective that software is a commodity when the product appears to the customer as simple and easy to use. Masking complexity from customers doesn't sound like much of a commodity, unless the pricing is extremely competitive.

For the vendors, the cost of hiding complexity can be high, which makes their cost to deliver the goods less than with a commoditised operation. In the end, the cost to deliver products and services is passed on to customers, so the winners are those products and services that are perceived as high value and at the best price.

Schwartz declared that commoditisation is fundamentally about competition and substitution. Mickos added that creating commodities takes "enormous skill, innovation and competition".

Schwartz's comment, however, rings most true. When you buy soap, (not SOAP) you have a huge number of choices and a competitive marketplace. Soap aficionados may find very granular differences among soaps, but for the majority of the market soap is a "commodity". You can easily substitute one bar of soap for another without penalty. You don't need to redesign a bathroom to choose a new soap. No vendor lock-in.

Similarly, if you want to buy a desktop or notebook computer, you have several choices and can easily substitute one for another. Are PC processors a commodity? With two manufacturers -- Intel and AMD -- there is some competition, but you can't simply substitute processors on motherboard. Not exactly a commodity, especially given Intel's market dominance.

Another facet of the commodity theory is that products are less differentiated. With open-source software and the growing standards movement, vendors have less opportunity to grandly differentiate or create customer lock-in. Does that make open-source software a commodity? Not exactly, if you use Linux distributions as an example. The differences among Linux distributions, for example, are just enough to be painful in making substitutions, as Schwartz likes to point in reference to differences between a Red Hat Linux distribution and a SuSE distribution. Even IBM's Rod Smith, vice president of Emerging Technologies and author of the letter asking Sun to open-source Java, said that binary compatibility between Linux distributions is "atrocious". And, Red Hat's market dominance allows the company to make decisions about pricing and product direction that diverge from commodity notions of substitution and competition.

One intractable problem is that innovation is often in conflict with concepts like compatibility, interoperability, substitution and the road to commoditisation. Sun and IBM, for example, have to find ways to innovate throughout the software stack to convince buyers that brand x is better than brand y, even if they have similar components, without compromising the capability to substitute one vendor's component for another.

According to IBM's Smith, commoditisation happens in different places in the software stack. "Customers get confused with interoperability and commodity. What functions do we build upon and provide greater value propositions?" he asked. Smith asserted that for customers, commodity is trust that they have more control over the people they work with. In an idealised IT world in which competition and substitution reign, customers will have more control and influence in the buying process.

In the real world, vendors make sure that their own components work well together, often at the expense of interoperability with competing vendors' offerings. This is the "open-proprietary" mode of operation. "Sun and IBM want to be in a unique position themselves with virtualisation solutions... from a customer standpoint that's not great," Smith said.

Roman Stanek, chief executive of Systinet, described the cycle that the IT industry undergoes as a tension between innovation and interoperability. "IT moves in steps. People interoperate, then innovate. Tension increases, and then a new level of interoperability [evolves]. In the next couple of years, there will be another standard to bring vendors together."

Web services are providing some level of commoditisation, in providing a broad framework for interoperability, substitution and competition. But, it will be many years before we have directories full of interchangeable Web service components. Even then, I find it difficult to believe that Web services are going to commoditise applications. Innovation and the desire to have some unique differentiation tend to trump commoditisation as defined strictly by substitution and competition. But, going back to the definition of a commodity as a product or service for which there is a universal or perpetual demand, IT fits the description from a macro view. The problem will continue to be that the IT industry is dynamic and constantly changing.

As Stanek said, IT moves in a step function. The constant innovation and competitive forces that have propelled the industry forward will continue to inhibit commoditisation in many parts of the software stack. However, standards will continue to evolve and take root, and the cost of continued innovation or differentiation at some parts of the stack won't be worthwhile. Don't expect a commodity data centre that you can buy from a vending machine soon, but there is at least recognition that part of reducing the complexity and cost of IT is based on substitution and competition.