Do we always need to know why?
Today's 'optimost' newsletter compares multi-variate testing first to American Idol and then NCAA 'March Madness'. It argues that the second comparison is more appropriate because of transparency and the wealth of information available about why Florida won its second straight national championship. American Idol is shrouded in mystery, although DialIdol and others make it more interesting.
From a testing perspective, March Madness is clearly champion-challenger or A/B testing. Two teams play, winner advances.
When it comes to peoples' tastes, prediction is difficult at best. In a study by Columbia University on teen downloading it was harder to predict the 'success' of a new song when the people could see the download rankings. Seems that there are a lot of interactions and some sort of gestalt that make books, movies, bands, etc. successful. Marketing is about leveraging and building momentum, not necessarily decomposing things into a billion parts.
Testing, and I highly recommend it, is not quite the same as measuring success. It is about improvement and validating decisions (assumptions.)
There are times when 'why' is important and there are times when it isn't.