In the course of those conversations, I was surprised by how differently we supposedly similar open source companies run our operations. We're each an open source company, but with varying licensing, sales, and support models. That's a good thing.
But it's also a perplexing thing if you're trying to weave together a common theme between them.
After our meeting, I spent some time on Sourceforge, pulling download data and correlating it to company revenues for these and other open source companies. After awhile, similarities started to emerge from the data.
Let's use JBoss as an example. Here are the company's downloads over the course of the project:

The red numbers underneath reflect revenue increases year over year. (All the numbers are available out on the web - you just have to dig.) Assuming Red Hat's CFO was telling the truth (and he wouldn't last long if he weren't), JBoss went from next to nothing in revenues (2001) to $60M. (Btw, this is pretty incredible for any company, much less one that started giving away its software only a few short years ago, and has stubbornly persisted in giving it away for free. Nice work, Marc.)
How? I took a walk through the Wayback Machine, to see remember what was happening over the past five years at JBoss, and see how the rising revenues correlate with downloads, product releases, etc.
The most intriguing thing I found is that there has never been any magic formula at JBoss (or at the other successful open source vendors). JBoss made great software. The media picked up on that and reported it. People heard about JBoss, and then downloaded JBoss' software. The biggest spikes in the downloads correspond to significant product releases (usually developer preview versions got the most traction upfront). The biggest spikes (or, rather, gradual but insistent rise) in revenues came 6-9 months after JBoss had started working with HP and Novell.
What I actually find almost shocking in the data is that JBoss' downloads have stayed relatively constant over time. It's not like they jumped 10X each year, or even 2X. They just slowly, incrementally grew. So, contrary to the argument that open source is a volume game, I'm coming to believe that while volume is important to get you through your initial hiccups and customer wins, the real measure of a successful open source company is its ability to convert whatever volume of downloads/would-be customers it has.
JBoss never had the tens of millions of downloads that Red Hat or MySQL have, but it was able to convert a significant percentage of its 50,000 - 100,000 downloads per month into paying customers.
How? Well, one way is simply through brand. The more momentum JBoss has enjoyed, the better its conversion rates have been. That's why the numbers from 2005 to 2006 rise so rapidly - it has established itself as the open source application server to beat. (You can actually trace the download activity to analyst reports and other news JBoss highlighted on its website through the Wayback Machine.)
The other equally important way is through push. That initial 3X JBoss enjoyed seems to stem, in significant part, from the added marketing Marc did back in early 2003 in the support services JBoss offered:
Knowledge from the source. JBG services are delivered directly by the developers of the code. JBoss Group consultants spend 50% of their time writing Free Software and 50% of their time consulting. This is a novel and unique service in the software industry.January 30, 2003. JBoss Landing Page.
Purchasing JBG services is a value added investment in your people and your IT infrastructure. It is transfer of real information from our teams to your teams. Spend your money on knowledge not empty licenses.
You can also see sales start to accelerate after Matrix, Accel, and Intel Capital invested $10M in JBoss back in February 2004. The company started to ramp its sales force, and revenues followed.
All of which confirms something I'll be presenting at OSCON. (It also, I think, confirms much of what Larry was arguing in his post about "natural revenue growth.") Below I've mocked up a (very) rough idea of what the appropriate sales model is for the number of downloads your company has:

The more volume, the less bandwidth (and need) you have to reach out in a direct way to your user base. The less volume, the more need to "push" sales to maximize conversion rates on the few downloads you have (all the while doing what JBoss did to increase downloads: build a fantastic product and try to get the media and analysts to notice so that people will know to download it and try it out).
The secret sauce, then, for JBoss and all successful open source companies is an exceptional product. Next, be sure to build momentum for your brand so that conversion rates (and sales) rise even as downloads level off. JBoss is the perfect example of a company that has done this exceptionally well, along with its new owner, Red Hat. In this way, creating a successful open source company really isn't dramatically different from building a strong closed-source company. The difference is in all open source's other benefits that give you unfair competitive advantage. But that's another blog entry.
I'm very surprised by the flat download numbers. I suppose this could be partially explained by the fact the JBoss now ships bundled in Linux distros, is distributed by 3rd parties (like Alfresco), and comes pre-installed at providers like Rackspace. The flat numbers could also be due to a more stable release cycle. But the chart you used also includes their new products like Hibernate so it really should be increasing. I really don't know how to explain it but it seems wrong to me.
ReplyDeleteI agree with pretty much everything you've said here but I think you are discounting the growth in the JBoss service offering. What JBoss really sells is support. And that service has become substantially more valuable since 2002. In 2002 the support offering was unproven and a confusing mix of direct email access to JBoss Group consultants and "CVS HEAD" fixes. In 2006, the support offering has a fantastic reference client list, a customer support portal, the JBoss Operations Network tool for managing servers and updates, and is packaged for easier purchase through a subscription.
IMHO, creating a compelling service offering has improved JBoss's conversion rate more than their increased marketing.
I'm looking forward to your session on Wednesday.