Sure, but if you measure that for a ten thousand people (controversially, ten thousand people that aren't already programmers may be better) you can get a good idea of what is efficient and what isn't.
Maybe, but also maybe not. Ten thousand people giving ten thousand uncorrelated responses is also a possible, not unlikely result of such a test. There's also the question of whether or not your methodology for "measuring" what is effective and what isn't is even possible, let alone sufficient for a definitive conclusion.
