The Science of Shopping (1996)
gladwell.com
gladwell.com
Other than this, I think the safest bet is to trust the data, e.g. A/B testing. Being too smart without any testing could lead to fatal mistakes - Google has thousands of phd's and they still trust real user testing more than "opinions" (at least according to Peter Norvig :)).
Paco Underhill seems to use the same method, i.e. observing real customers and then drawing conclusions.