Flawed data is sometimes worse than no data at all. Data with sampling bias, too small a sample size, lack of double-blind rigor, etc.
Strictly speaking, "surveys" are only really used by the softer sciences, like sociology and psychology, and both fields agree it's the least rigorous (albeit easiest) method of accumulating data.
Corollary to this, a little bit of flawed data can sometimes give a good idea in the general direction, but more often than not it's worthless because it shows false correlations, leads to erroneous conclusions, stimulates debate in the wrong area, questions prior evidence, causes one to hypothesize irrelevant answers, etc.
Boy, talk about "the perfect is the enemy of the good"!
Is salary negotiation a branch of particle physics now?
Yes, this survey is much less valid than if we could magically collect all salary data for all developers in the area. But I don't have that magic wand, and this survey is far more useful to me than its closest competitor, which is "a buddy moved to the Bay area a few years ago and told me what he'd been offered".
Strictly speaking, "surveys" are only really used by the softer sciences, like sociology and psychology, and both fields agree it's the least rigorous (albeit easiest) method of accumulating data.
Corollary to this, a little bit of flawed data can sometimes give a good idea in the general direction, but more often than not it's worthless because it shows false correlations, leads to erroneous conclusions, stimulates debate in the wrong area, questions prior evidence, causes one to hypothesize irrelevant answers, etc.