Flash! Some OSS business models don't work

For more than a decade, b-school professors have been teaching about business models. However, I suspect sometimes we spend too much time emphasizing how important it is to have or understand a business model, and perhaps not enough about the importance of a good business model. (Regular readers know that I often write about bad business models, broken business models and even the decades-long search for a viable business model).

One important area of business model experimentation has been open source business models. This morning, my friend Matt Asay posted about Monty Widenius, who five years ago sold MySQL to Sun Microsystems for $1 billion, and then bailed out of Sun when the acquisition helped drag Sun down (or failed to lift it up) — leading to Sun’s eventual dismemberment by its new owner.

His opening paragraph sets the stage:

Monty Widenius, co-founder of MySQL and founder of MariaDB, just came to a surprise revelation: most people use open source for free. What's so surprising, however, is not this fact, but the idea that Widenius wouldn't have learned this 13 years ago when he first released MySQL under the GNU General Public License (GPL) and his company's revenues dropped 80%. The lesson here, however, isn't that there's no money in open-source software, but rather that some strategies for monetizing open source are effective, while others are not.

As Matt notes, it’s inherent in offering real open source (i.e. something that conforms to the Open Source Definition) that people can and will use it for free. Widenius' current suggestion is to have semi-open source (called “Business Source”) which might generate revenue but can’t be called “open source.”

I spent almost a decade researching how firms make money off of open source, including one of my favorite (and most-cited) unpublished papers. During that study, it was obvious pretty quickly that you had to give away something useful — so people would adopt your software — and hold back something valuable (e.g. support, add-on modules) so that people would pay you something. The balance was made tricky by the one-way nature of open disclosure of source code: once something was free (as in beer or speech) it was technically and legally hard to make it un-free.

Matt barely contains his snark as he cites all the people who’ve long since figured this out:

Jim Jagielski, president and co-founder of the Apache Software Foundation, suggests that "if your open source project isn't successful with FOSS licensing, it's not the license's fault." Rather, it's a matter of trying to charge for the wrong things:

what's "destroying" open source isn't people not paying for it, but wrong ideas on WHAT they should be paying for

— Jim Jagielski (@jimjag) May 30, 2013

To wit, Facebook, Google, Amazon and others make billions of dollars selling services around open-source infrastructure, while Red Hat mints over a billion dollars annually selling a certified, binary distribution of community-developed Linux. There is plenty of money in and around open-source software. The open-source license doesn't prevent this. It enables this.

Widenius is a smart person. He'll figure it out. It's only surprising that his experience at MySQL didn't already teach him this lesson.

I’m guessing that Widenius has already figured that out, but he can’t see how to get there from here because he’s painted himself into a corner.

(BTW, there’s always been a difference of OSS business models between firms that are givers and takers. A low cost, low risk strategy has been to be a taker — or net taker — by leveraging OSS created by others. It’s the givers that incur the cost of R&D for their free stuff, and thus bear a sizable risk of not being able to monetize well enough to cover that cost.)

One thing I haven’t seen mentioned is the inherent scale requirement for OSS and other freemium business models. As my research continued, it became clear that a revenue model with 1% (or 10% or 0.1%) conversion rate might work with millions of users, but the identical strategy would fail if you only had thousands of users. So copying Red Hat (or MySQL or Sleepycat) was going to fail miserably unless you had a smash hit that would support a 1% conversion rate. This is something that has recently become painfully obvious to all the iPhone game makers who used free games to win adoption and hoped to sell upgrades.

So yes, market leading companies with billions of customers can afford to give away software and monetize off of some small fraction of their customer base. But for new companies trying to launch now — even those run by smart experienced OSS veterans — an OSS (i.e., freemium) business model is at best a risky gambit. At worse, it’s a fools’ errand, because it’s no more feasible to create the next Red Hat or MySQL than it is to create the next IBM, Cisco or Google.