Manufacturing Software

Obviously the software economy is very different from that of other businesses. The most widely-acknowledged difference is in the split between overhead and marginal costs: while prices for many physical goods have historically been dictated by the costs of production, producing one extra copy of a software product is effectively free.

The “zero marginal cost” property of software is often misinterpreted, however, as suggesting that there is no software analog to manufacturing. And I think this misunderstanding leads people to miss one of the other unique properties of software.

Whether you’re producing hairdryers or hand-carved statues, the general workflow is the same: first come up with a design, then get a bunch of people and/or machines to perform the construction, sell the finished product, and repeat the last two steps for each customer. In some businesses the design stage alone is incredibly expensive—automobiles, for example. In others, there’s only one manufacturing run—e.g. constructing an office building.

The important point is that in traditional businesses there’s usually a pretty clear separation between design and manufacturing. Once the product has been sold, the “intellectual property” produced during the design is out for all the world to see. In some cases IP law may hinder others from copying your design for a while, but in general anybody else will be able to go out and manufacture a knock-off for something comparable to your own cost of construction.

In software, however, there isn’t a clear line between “design” and “manufacturing”: software is effectively a design that’s been completely rigorously specified; all the detail work and quality assurance that is typically a part of the manufacturing process is performed centrally. Once a software product has been released, competitors can take a look at the list of features (and possibly even poke around to see how bits of the implementation work), but translating this “design” into something that can be produced near the zero marginal cost still requires a huge investment.

Other businesses already have some of these properties. Setting up a manufacturing process for a new product can be very expensive and may only be practical for large volumes. But such costs still exist at the margins: producing twice as many products requires similar setup costs. The cost of “designing” the production process itself is a rounding error in traditional businesses.

The major impact from this difference between software and other products stems from the fact that manufacturing can be continuous for software: once you’ve finished manufacturing one product you can build on what you’ve already got to produce the next version: the manufacturing investment is (mostly) cumulative. This makes early leads in software much more valuable than leads in other fields, even where market share is irrelevant. Other companies may be able to invest and eventually catch up with Apple’s iPhone/iPad software (Palm’s WebOS and Google’s Android are both getting there), but if everyone keeps running at the same speed Apple will remain well out in front until the weight of their manufacturing legacy weighs them down. Waiting until a market is mature and then leapfrogging the competition is impractical in markets where manufacturing doesn’t happen at the margins. Sony take note.