Facets are a technique in modern IR systems. In essence they  replace classical deep hierarchy classification scheme with co-current classification using several shallower classification schemes that operate orthogonally on the subjects.

Facets are integrated into the interface of web sites like eBay and zap.
These web sites have used usability studies to adjust their user interfaces to work with facets. They now offer a Superior  experience for their users.

Searching using facets allows to specify search parameters from different classification hierarchy. As more parameters are added potential results are filtered out. Specialized parameters appear as they become relevant to apply more fine grained filtering. But  Facets can help the searcher to avoid dead end with no results.

Unlike the black box presented by classical IR systems using facets is more like an exploration of the classification space. The users can rapidly assimilate shallow hierarchy of concept (3 or four levels deep) Which is simply not feasible for a deep hierarchy (12 or more levels deep -- with all the top levels acting a abstract categories of non indexed terms.

For an information architect Facets can offer added advantages. One can derive Metrics based on Faceted classification to determine what region of search space used by the users cluster too many results, and which offer to few. Once these regions are identified the classification can be adjusted.

Where there are too many results an additional may be the natural solution to separate the results into new buckets. Where the reverse is the case one would set the system to not offer as many facets specializations as the result clusters become too small.