I’m getting really tired of pundits claiming that big companies can’t innovate, and that startups are the saviors of our economic future.    I’ve worked in/with big companies and startups, and have seen tons of innovation and many brilliant people in both. The biggest difference is a simple and obvious one – big companies become conservative because they have something to conserve, and startups take big risks because they have nothing to lose. Their respective shareholders expect these behaviors, and this difference in operating model – not lack of innovation or smart people in big companies – explains the difference in outcomes. Otherwise said, big companies don’t bet the farm, simply because they have a lot more mouths to feed than startups do. 

The latest article of this genre that set me off was in Silicon Valley Insider.   Henry Blodget pimps a video of Kevin Ryan talking about “Why big rich companies can’t crush tiny startups“.   I posted a long comment that focused primarily on the fundamental logical flaw in the kind of reasoning Ryan (and many others) use, and thought I’d share it here also. 

Ryan’s “analysis” suffers from the same kind of selection bias that makes mutual fund returns look better than they really are. All the failed mutual funds – or startups that were killed by big companies and forgotten – aren’t taken into consideration.
By only studying successful cases – e.g. the Googles or Yahoo!s, etc. – you essentially guarantee that you can prove the hypothesis that big companies can’t do the things (selected successful) startups do.
Another way to appreciate this is to examine the fundamental flaw of books/articles that try and tell you how to be a winner by studying/extracting the traits of winners. Some authors tell us, for example, that successful startups are run by risk takers, therefore taking risks is clearly the way to be a successful entrepreneur.  But – if the authors bothered to look at failed companies – they’d see that those entrepreneurs were also (surprise) risk takers.  So we haven’t really learned anything from the “analysis”.  Statistically speaking, that’s because you can’t just select a sample, extract a characteristic and call it significant – you have to compare it to the characteristics of the population to see if there are in fact differences from the baseline (or base rate).
These lightweight Gladwellian types of analyses that focus on selected stories that support the hypothesis “feel good”, so they are easy to sell.  But it takes a lot more work to do it credibly. And even the guys who do the hard work – like Jim Collins – have selection bias issues when they try and track/compare companies (and company types) over time.