The effectiveness of shallow hierarchies for document stores

Employees spend as much as 4.4 hours every week searching for documents that they never find. Despite this cost, most managers continue to believe that there is no viable alternative to keyword search. In this paper we present the results of an experiment which uses the eight level hierarchy of ABI/Inform to test how many levels are necessary to retrieve one specific paper. Our findings demonstrate empirically that a browsable subject hierarchy of just four levels provides almost as accurate a search result at deeper layers. Therefore the cost of implementing and maintaining a browsable hierarchy is not nearly as high as is frequently estimated. This has significant implications for both researchers and practitioners.