This is clearly his opinion, based on experience. There isn't really any good way to collect data on why a company fails. So basically by definition, it is impossible to do an analysis that would provide you with a citation.

... is it? I figure this is one of those things that you can either take at face value -- or not.

The real question is did you find the rest of his information useless, if you did, then does it matter if he has an actual statistic here? If you didn't find it useful, would the citation make a difference at all?

I'm just saying it seems like a weird thing to focus on in the scope of this entire article. It probably wouldn't destroy the entire reading experience if you just mentally said "He probably meant 'Many'..." and kept reading to see what was important in the article.

I'd assume this is correct. Given that most startups end up launching with something, wouldn't it make sense that those that fail do so for reasons outside of their ability to build a product?