I've always been pretty skeptical of the 90% number, but I don't think this article clears up the matter. I'm assuming all of us here aren't interested in opening up a mini-mart (rather we're interested in startups in the PG definition of the word -- basically something that scales). And yet, the "over half" figure that's quoted in the article I'm sure includes things like mini-marts. And I would guess that a mini-mart business has a better chance of staying open over a year than a tech startup does. So the "over half" might be inflated in the other direction.
Also, I wonder if companies that in reality fail before one year might just tend to wait until the next tax year comes before officially shutting it down with the government. If so, depending on how the stats were found, this could inflate the success percentage.
There was some study - I think it was linked to on Dharmesh Shah's blog (onstartups.com) a month or so back - that indicated the success rate for first time entrepreneurs in VC-funded tech startups was 18%. For entrepreneurs with a previous failure it was 20%, and for entrepreneurs with a previous success it was 33%.
I suspect that the success rate is somewhat higher for bootstrapped tech startups, because you're given the option of failing and trying again with the same company, different idea. Usually if you fail with a VC's money, they liquidate the company. If you fail with your own, you can take the assets you have and try to redirect them to a new line of business.
I thought I saw another study somewhere that said that the average number of failures a bootstrapped startup goes through is 4. That squares with the success rates for VC-funded startups above, if you assume that each failure simply means the entrepreneur tries a new idea. You have a 18% chance of succeeding on the first try, a 33% chance of succeeding by the second try (82% * 82% = 67.24% chance of failing), a 45% chance of succeeding by the third try, and a 55% chance of succeeding by the fourth try, for an average of 3-4 attempts necessary before success.
If the success rate for VC funded startups is 18%, then I could easily believe the overall rate is 10% or lower, because angels fund more startups than VCs, and they are less selective.
More selective doesn't necessarily mean a higher success rate, if VCs can't pick the winners at a higher rate than their general frequency in the population. Wasn't there just an article on here about how VC-funded "A teams" do no better, and often worse, than VC-funded "B teams". And I thought your "Unified Theory of VC suckage" mentioned this too, though rereading it, it seems more about how VCs take winners startups and make them fail rather than how winners avoid the VCs to start with.
That said, my high school math teacher and first employer was an angel investor, and he said his success rate was *well* under 10%.
A lot also depends on your definition of failure. When Ludicorp shut down Game Neverending and launched Flickr, would that count as a failure? If they had taken millions in VC funding, would they have been able to shut down GNE and say "We're going to build a photo-IM client"? I'm currently working for a (profitable) startup with 4 products. Only one of them is generating any revenue. If that had been VC-funded, it likely would have been 4 separate companies, 3 of which were failures.
Also, I wonder if companies that in reality fail before one year might just tend to wait until the next tax year comes before officially shutting it down with the government. If so, depending on how the stats were found, this could inflate the success percentage.