Until that happens, the implicit assumptions is that the startup tried shopping itself around to first-tier VCs and angels, failed, tried to go to second-tier VCs and angels, failed, so crowdfunding is their last resort.
Alternatively, a crowdfunded startup scorned by VCs needs to make a really nice exit or achieve a very impressive unicorn valuation.
Figuring out a way to create that VC/Angel connection implies abilities and characteristics that they pattern match to past success. There may be high potential teams that cannot pass this filter but will still succeed, but the industry's deal flow is large enough that they are fine with a small number of false negatives since false positives in their portfolio hurt more, especially when you consider VC math needing 100x-1000x return potential from each of their companies.
It's just like the FaceGooAmaSoft hiring filter we're familiar with that is designed to prioritize low false positives at the cost of having a higher false negative. (skilled programmers who aren't interested in the contrived coding quizzes or can't do them well will be a false negative in the hiring process)
This is backwards. Funding a company that goes nowhere is far smaller an error than turning down the next SpaceX.
Afterall didn't A16Z catch some criticism for not investing a large enough sum into Instagram (which iirc had 500x returns or something) to make a large enough difference to the fund?
A costly adherence to their own ethical rules, but it reinforces how seriously Andreessen Horowitz takes portfolio conflict.
source: https://bits.blogs.nytimes.com/2012/04/20/how-andreessen-hor...
Biases like this are market inefficiencies that can provide outsize returns for those who learn to look past it.