Facets + Flash
Last year I wrote about how browsing interfaces are better than search/combine interfaces for faceted information. Most search/combine interfaces don't expose the data, meaning that it's often too easy to dial up an empty address (a combination of facet terms for which no resource exists). The Wine.com "power" search is one example of this; here's another.
Assuming every facet term is used at least once, the minimum number of resources in a simple faceted information space should be equal to the number of terms in the facet with the most terms. (I say "should be" because I'm too lazy to do the math.)
On the other hand, a faceted information space can have a huge number of addresses--five facets with 10 terms each equals 100,000 possible addresses. But that same "space" need only have 10 resources to ensure that each term is used at least once. Which means as few as .01% of the addresses could be populated. Interestingly, that's also the probability of a user finding one of your ten resources by combining those facet terms through five drop-down menus. Put another way, that user is more likely to die in a car accident this year than find one of those resources.
The camera finder, though not perfect, is a great way of exposing the facets and the data simultaneously.
