I'm at the O'Reilly's Strata conference in NYC, where a NYU professor shared his research on consumer purchasing behavior. He asserts that 'What people "say" is different from what they actually "do"', so he recommends conducting behavioral studies by measuring what people "do", not what they "say".
He used camera buying from Amazon as an example, where Zoom is talked about a lot as a feature but has lower impact on purchasing behavior. Whereas Battery Life and Megapixels are talked about less but have higher impact on purchasing behavior.
We often conduct such analysis using surveys in enterprises. To do what he suggests will be more accurate but will cost more... it is much more costly to follow what people "do".