David Heinemeier at 37signals published a post titled “Bug tracking isn’t a network-effect business" in response to a piece by Joel Spolsky in Inc. asking “Does Slow Growth Equal Death?” Heinemeier’s post is a possibly intentional misreading of Spolky’s so as to make it seem more controversial than it might otherwise be, with sentences like “I don’t think Spolsky notices a difference.” So here is why this is a non-debate: Spolsky is concerned about the impact of scale economies (in fact the word “network” never appears in his article nor any direct allusion to it). And he is right to point out that scale economies can make a huge difference in software. Because the cost of making software (whether installed or SaaS) is mostly fixed, any incremental sale provides 80%+ of gross margin. That means that once you have covered your fixed cost you have a lot of “play money” to start throwing around. You can, like Spolsky’s Oracle example, spend heavily on marketing (and unless you believe that marketing is 100% useless) that will make a difference to sales. And of course here is where things start to get interesting, because you can have such a strong feedback loop that it allows the first player that gets to scale to become 10x bigger than the next player (grow faster, make more money, spend more on marketing, grow even faster).
This is something important to worry about for software companies and Joel is right to be thinking about it. Now what about Heinemeier’s post. Is it really true that bug tracking is not a network effects business? In the strongest form of network effects (e.g. facebook, linkedin), the answer is no. But that doesn’t mean anyone offering bug tracking should not be thinking about how to create some weaker type of network effect. By that I mean doing things so that as more folks use the software or service its value to each user does increase gradually. It seems to me that there are at least two different ways of doing that:
- The first would be to start gathering statistics and sharing those in the network. For instance, how does your bug severity distribution compare to others? How about your resolution times? This is information that gets better with more people using the system. Not only will it be statistically more relevant, but with more users you can also offer additional granularity (e.g. compare your bug closing rates to similar companies, not just the overall average).
- The second opportunity for creating weak network effects is some kind of plug-in or extension architecture. As more people use the bug tracker, some of them will come up with neat extensions. With a plug-in architecture they can make those available to others. Now that makes the system more valuable to everyone who has the same need, which in turn is likely to attract more users, among whom there will be future contributors of extensions.
So I believe the assertion that bug tracking is definitely not a network effect business is wrong. The effects may be weaker, but anybody offering software today and not thinking about how to create and sustain network effects in what they are doing is missing a big opportunity.