One of the best examples of this problem is something entirely outside of software. Coke and Pepsi. It's well known that people prefer Pepsi to Coke in blind ta

One of the best examples of this problem is something entirely outside of software. Coke and Pepsi. It's well known that people prefer Pepsi to Coke in blind taste tests by a very wide margin. Yet in the market Coke outsells Pepsi by a very wide margin. This anecdote was used for quite some time as an example of the power of advertising. But it turns out the truth is something much more simple. While people do prefer Pepsi in blind taste tests, that's because in blind taste tests you're only getting an isolated sample of the product. To cut out the software metaphor, people prefer Coke when it comes down to drinking an entire bottle of the product. It has nothing at all to do with advertising or false preference - Pepsi was just testing the wrong thing. The point here is that just because people think they want or like something doesn't necessarily really mean they do, especially when what they're 'consuming' is in some way different than the regular long term consumption of the product (as you hit on). I don't think it's much of a coincidence that some of the most famous marketers, most relevantly here probably being Steve Jobs, weren't particularly interested in what the consumer thought they wanted.