One thing that's permanently changed the way I answer the question "what do people want" is this video, about how one man discovered that people are segmented,

One thing that's permanently changed the way I answer the question "what do people want" is this video, about how one man discovered that people are segmented, and they won't tell you what they want: http://www.ted.com/index.php/talks/malcolm_gladwell_on_spaghetti_sauce.html http://www.ted.com/index.php/talks/malcolm_gladwell_on_spagh... What you should end up taking away from it is that there is no "perfect" thing you can offer, but rather several different types that appeal to different segments of the population. And these segments won't come out and say: "I want ____!!" Using the video's example, when asked what type of coffee people want, nearly everyone would say "dark rich hearty blend" (or something to that affect). But what they actually prefer is "milky wheat" coffee. This wasn't really known until taste tests were done asking people to rate their satisfaction.. and it turns out there is a strong correlation into these segments. Granted, this is all based on food products, but it applies to the web as well. Jacob Nielson discovered years ago that people are divided into major groups of searchers and browsers, for finding content. I bet that there are tons of other segmentations for user interface type things that I just don't know about (it's not my specialty), as well as expectations for features. For instance, maybe your users would actually prefer it if your links were bold, underlined, and blue, and that when you hovered the underline goes away and it turns a lighter color. Or maybe they'd actually like some sort of RSS feature (that didn't require them to know what RSS was). But they probably won't ask for it. I don't understand your logic... users know what they don't want... add in some magic.. and suddenly they will express their desires? If that magic is the process of deduction, good luck trying every permutation, and seeing if users don't complain. I think A/B testing works best to decide between two implementations. And I think tossing in a new feature and asking: "do you like this" works just as well.