Open Source Business Practices and Conversion Rate Myths

[Update: 3 March 2009, 12:13] Added some more ideas here.

With the Sun Microsystem recent move to acquire MySQL AB, open source business models will be a topic of much discussion again. MySQL AB, like Red Hat, has always been one of the examples everyone points to for how an open source business should be run. One of the oft quoted statistics of the MySQL business is "one customer for every thousand users". This number is then quickly put into context as "probably too big" because MySQL is available in so many places that trying to count downloads and users becomes impossible. When JBoss was acquired by Red Hat, the publicly acknowledged conversion rates were 3% (JBoss) and 10% (Red Hat). People start making assumptions about business models based on driving downloads and user community size. And that's where the problem starts.

That's an order of magnitude difference. In a bricks-and-mortar world, business differences measured in a few percentage points are spectacular when comparing company ratios. In a digital world, where the cost of goods sold and marginal cost differences change, it doesn't seem right that we would be seeing orders of magnitude differences between companies for this type of ratio.

I participated in a recent informal (qualitative) survey around purchased support for Linux systems. [It was definitely NOT quantitatively or statistically significant. But anecdotal evidence can be interesting.] In operations of 100 servers or fewer, no one even considered buying support from their vendor for ANY of the servers. In operations of greater than 1000 servers, no one questioned buying support for ALL of their servers. Even "small" operations of 100 servers were still mission critical.

Somewhere in the order of magnitude difference in the server farm size, buying priorities change IT budget allocations. Maybe it has something to do with relative percentages of the budget spent on hardware/software/support versus headcount/expertise. Or with the cost of downtime to the business for a particular application environment. (A friend with a "small" server farm of ~100 servers told me recently they didn't hesitate to pay for JBoss support while migrating a critical application from a BEA Weblogic environment, but didn't consider paying support on the Linux servers themselves.) Or maybe it has something to do with the relative margins of the business. But I'm pretty sure the change in buying habit has more to do with the customers' business needs than the vendors' distribution practices.

Regular readers of this blog have heard me rail against the idea that open source is some sort of different business model. (Again, I'm not talking about collaborative software development in community here, but rather businesses that use such software.) It's definitely an area we need to go think about some more, figuring which software business ratios are significant and which practices to encourage.

Comments

[Niraj J]

Good post. Could not agree more with "Regular readers of this blog have heard me rail against the idea that open source is some sort of different business model".

I think the only difference is that open source is a very different market capture strategy (to me the next level of Microsoft's tolerating Piracy strategy).

Another important aspect is knowing what kind of open source business you are in. There are commercial players and there are not for profit players(example Apache) - both have different models around running a business

some more perspectives on the topic
1.http://www.gandalf-lab.com/blog/2007/04/does-open-source-make-ecomonic-sense.html
2.http://www.gandalf-lab.com/blog/labels/open%20source.html

[Michael Tiemann]

Stephen, I think you have the right handle and the right sense here, namely that the larger the company, the more likely they are to consolidate support and buy from a qualified set of 3rd parties. This is precisely the dynamic that I have seen all over the world: the "early adopters" don't actually adopt support, just source code. But when the code is handed over to "mainstream" users, they want support because they know how to build and run a shop with defined support models.

I could swear that last year I came across a university-led survey that asked this same question, and concluded the same thing: that smaller outfits are more likely to self-support, whereas larger outfits are more likely to contract support. This may explain why building an open source business is so difficult: there just isn't a workable revenue model until you have enough scale, and you cannot get that scale until you prove to the mainstream customer, over a period of years, that they can trust you to be smart with their money.

That does not make open soure a poor business model, just a very, very defensible one. And it may explain why open source companies like MySQL and Trolltech are much more likely to be acquired by a company that does have scale than to attract market capital to become a company with scale.