Aren't all these sites limited by (and biased by) the selection of offer data they collect? Unless they are working directly with payroll companies to obtain a uniformly random sampling of offers, they're not going to be accurate. What incentivizes employees who get below-average offers and companies who give below-average offers to share that data with these salary sites?

Example: Everyone who even knows someone who makes $500K at Facebook seems to love posting about it here, but nobody who makes $100K there seems to feel the need to boast about it.

That's a reasonable distinction but just look at the data itself.

There are tons of low level employees posting their salaries. If you work in those roles, that data could be more valuable to a candidate than Glassdoor's amorphous projections.

I think the user behavior that facilitates this is the same behavior you see represented on Blind.

Levels allows HR departments to post their actual salary bands, which they're normally more than happy to do. So you're not just relying on user reported data. They're also incentivized to do it because they want compensation benchmarking information (big companies want to know how much their competitors are paying so they can calibrate their own bands to market rate).

https://www.levels.fyi/offerings/

They also have users upload W2s, offer letters, and supporting documents in addition to proof of identity.

https://www.levels.fyi/verified/

They have their own job advertisement business and they partner with third party recruiting companies to get anonymous data about offers candidates receive.

They also just take having accurate compensation information seriously so they do things like outlier and forgery detection as well as seeming to have a much better algorithm for processing the information they receive.