The Economics of Software, redux

So there has been quite a bit of reaction to my thoughts on the economics of software. While most of the reaction has been quite positive, several economists have taken issue with my analysis. (Not surprisingly, I suppose; I’m sure I would take issue with any software written by an economist.) For example, Susan Stewart brought up the legitimate point about my conflation of macro- and microeconomic analysis, especially with regard to the supply curve for the firm. Susan is right on this, but we agreed that it doesn’t change my overall analysis. While feedback like Susan’s was quite helpful, that of another economist was decidedly less so: he was foaming at the mouth because I had the gall to ignore the theories that he set forth in his dissertation. His (largely ad hominem) attack revealed that (a) he didn’t read past the first paragraph or so of my blog entry and (b) that he knows absolutely nothing about software.¹ I ultimately got him to agree to (b), after which point I stopped trying to convince him of anything else. And at any rate, these qualms from the academy didn’t have anything to do with the crux of my thesis: that demand for a particular software product is largely price inelastic, that software vendors act as natural monopolists, that open source is an effective way of driving demand for complementary goods, and that this all adds up to a powerful supply-side open source movement.

From my peers in the software industry, the reaction has been more positive – and certainly much better informed. David Ogren had a thoughtful follow-up exploring the genesis for tiered pricing in software, and then another discussing the economics of software support. I view software and support as two different products, so I don’t think that David’s analysis invalidates any of my own. Paul Brown added his thoughts on how the economics of software and the economics of software support relate to one another. And John Mitchell was apparently inspired to ask an intriguing question: “what do examples of post-fiat economics have to do with open source?” I have absolutely no idea, but if I start getting paid in Papiermarks, there’s going to be big trouble.

The most common point of disagreement from those in the industry has been that I didn’t include the cost of sales and marketing software in the cost of software. At the risk of stating a tautology, the cost of marketing a product is not a part of the variable cost of making the product; the variable cost is simply the cost of manufacturing one more unit, and (for software at least) that cost is damn near zero. (The cost of bandwidth is the only argument to be had, and that’s pretty cheap.) This isn’t to say that sales and marketing aren’t necessary to ship a product, just that they don’t factor into the variable cost.

Another point of disagreement is my contention that software doesn’t wear out. I’m not budging on this one, and I refer doubters to my recent experiments with VENIX, software that has had zero maintenance applied to it in the last twenty years. The software works exactly today as it did twenty years ago – it’s as good as new.² Some mistake software that stops working as having “worn out”, but this is an incorrect inference: the software itself has not changed; a heretofore unknown attribute has merely been exposed – an attribute that was there all along.

I’ve been meaning to do this round-up of reaction for a while, but I was prompted by this tale of one database customer and the Firebird database. This vividly shows pretty much everything I described: a software company, seeking to extract additional revenue, jacks the price of the software that the customer already has by using one of the ugliest tricks in the book: the software audit. But the software company gets a bit too greedy, and overplays its hand by raising the price four-fold (!) – pushing the customer above their FYO point. (As an aside, I obviously stand by my original nomenclature.) Meanwhile, a hard-on-its-luck also-ran software product written ages ago ends up in the hands of a company that isn’t making much money off of it. Of course, it costs them nothing to manufacture it, so the company decides to give it away for free by open sourcing it. After that initial supply-side push, the demand-side takes off: development begins to blossom, with the demand-side adding many long sought-after features. And in particular, a compatibility layer is added that dramatically lowers the FYO point of everyone running the proprietary database. The pissed customer finds the open source product – and the proprietary product gets the boot (and a lawsuit, I might add). Everyone’s a winner. Or rather, almost everyone…

Despite Larry Ellison’s (expensive) attempt to fulfill his own prophesies, this is the true future of the software business: it may take years – and it may even take decades – but open source from the supply-side coupled with energy from the demand-side will ultimately drive the FYO point towards zero for many existing software products. Companies that can use open source to drive complementary goods will survive and thrive; those that can’t will slowly whither away, until one day their software is acquired for pennies-on-the-dollar by some company that just happens to have some complementary goods to sell…