I'd note that this bias extends beyond startups; see http://jeps.efpsa.org/blog/2012/06/01/falsification-of-previous-results/ http://jeps.efpsa.org/blog/2012/06

I'd note that this bias extends beyond startups; see http://jeps.efpsa.org/blog/2012/06/01/falsification-of-previous-results/ http://jeps.efpsa.org/blog/2012/06/01/falsification-of-previ... studies of failures don't just reveal things that increase failure rates, they prevent false positives. Imagine X is some innovation that is being talked up because recently several companies had or used X to succeed. If 19 out of 20 times X makes a startup more likely to fail or if X has no meaningful effect on failure rates at all, or if X only helps in situation Y, but you only hear about the times that people succeeded with X, you might try to use X, to no benefit or to actual harm to your business. If startup success is rare, then trying to generalize only from instances of it may lead to errors. We have more instances of failures, therefore we can learn a great deal about commonalities in startups that fail. Of course the smart thing to do is to make a hypothesis about X, and compare startups with X that failed and succeeded to startups without X that failed and succeeded to see if there is a significant effect on the rate. I'd note that errors of judgment about what causes startups to do well or fail get fed into the investor system. If X is useless or harmful, but your investors think it's a good idea, you will have problems.