Facebook wasn't "build first" in any appreciable sense. The prototype features were outlined in a well received editorial in the Harvard Crimson in December of 2003.

There was widespread outcry for the administration to build what Zuckerberg realized he could build faster.

Further, there were existing prototypes in house facebooks.

Source: user number under 10k

All of those, including Google are B2C companies at its core(I know the first 3 sell ads, but they'd be useless without consumers).

If you're selling to consumers, you need to know what they want, because asking them won't yield useful answers. Who would have told you "I want a text box where I can type 140 characters about how I feel" 6 years ago?

If you're selling to businesses however, you need to do customer and market research.

Really depends on the type of product, if you know it's something that the whole world would benefit from then maybe build first makes since, but if your targeting a niche market then for sure you want to get a barometer if your product is going to be needed and what the biggest pain points of those who would use it are.

For every five or ten successful build-first companies, there are 10,000 failures. Please keep things in perspective.

That is a dangerous conclusion - you can blow off just about any piece of advice with it.

What are the historical estimates of success for the sell-first case? Having both figures is necessary for perspective.

and how many build first and solve a problem no one has?

Lots, but at the end of the day it shouldn't mean that nobody should build first.

that's why a lot of products fail. They are solving problems nobody has.

People who, unlike a successful drug dealer, actually want to use their own product? I agree that if you are someone with desires not shared by the rest of society, you should probably go and talk to people and see what they desire...but it's also quite possible to think of something you and maybe some of your friends would use and have that be the guiding force in development. The problem with asking other people whether they'd like it, well [insert apocryphal story about Henry Ford and people wanting faster horses].

It's not either-or, it can be a mix. But it's not beyond the capacity of some hackers to hack on something they care about and, at the same time, have it be something that others will care about, unless all they're doing is trying to make a quick buck in the startup rush.