Do tech incubators help startups?
A recent Forbes article throws some heavy shade at the thousands of institutions that are trying to help new companies grow. The social media link sub headline is “New research from the Kaufmann Foundation shows that business incubators may not be helping startups as much as they claim to be.”
First of all, that really isn’t what the Kauffman report found. Their MAIN argument is that not enough work has been done comparing incubated companies versus non-incubated, which is an excellent point. There is a SINGLE study that did that, and the data is not encouraging. But it is only one study, and much more work is needed.
And that’s not going to be easy. To begin with, we need to not confuse incubators with accelerators: they do different things, usually have different sponsors, participants have different expectations, and measuring success is likely to be different too. Next, there are multiple confounding factors that make evaluating the impact of incubators problematic.
This is a very ‘political’ question, even within the tech community. Those who skew more free enterprise believe that bootstrapping and succeeding on your own are the ONLY paths to success, they talk about Apple and Facebook, and the fact that almost all tech mega-successes have NOT come from incubators or accelerators. These types tend to be opposed to any kind of dirigiste industrial policy. And many incubators are sponsored by governments.
There may be a selection bias. Companies that are very early hyper successes often attract significant capital early in their life cycle: as a result they never end up needing the support from an incubator or accelerator. This kind of “cream skimming” may mean that the pool of incubated companies is INHERENTLY less probable to end up as successes. The incubators aren’t the problem, it is the pool of candidates who NEED to use incubators that are statistically less likely to succeed. It doesn’t mean that helping them is a bad idea. One of the big challenges is that robust analyses of the incubators globally may show that they are not helpful on average. That wouldn’t surprise me: I suspect some incubators are good, and others are less good. A more interesting question would be “are there any incubators or accelerators that show a consistent pattern of helping start-ups become successful, and if so, what can we learn from them?”
The real problem then becomes the lack of good statistical universe. Let’s say the average incubator helps 25 companies per year, with 2/25 becoming a big success. If one incubator has three successes three years in a row, while a second incubator has only one success per year for three years, it would be tempting to examine the first for best practices, while dismissing the second as a failure. Bad idea: the difference between one and three over such a short period could easily be due to random chance. This is like picking your mutual fund on the basis of performance. Just because a portfolio manager picked the right stocks for the last five years doesn’t mean she will do it for the next five.
BIGGEST PROBLEM: what the f*** do we mean by success? What do we want incubators to do? Is it to turn companies into billion dollar valuations? Or billion dollar revenues? Create jobs? Get academic ideas into products? Employ university graduates? Build a tech ecosystem or critical mass? Backstop a critical skills area? Fill up an abandoned steel foundry that would otherwise be empty? (That last is a real thing.)
What if incubators are shown to produce identical outcomes to non-incubated companies…but allow management and founders to grow to the same level WITHOUT as much dilution through raising capital? That would certainly be seen as a benefit by the founders, who in turn might be worth $2 billion instead of $200 million on the IPO, and then set up a billion dollar research foundation in the city where the incubator was. That feels like a win to me, but how do you measure that sort of thing across the globe, and over short time periods?
I am sure there are more issues, but this feels like some useful starting points.