In a previous post I suggested that at a certain size hierarchies serve little purpose, actually this is not strictly speaking true. Hierarchies of data have two effects they both simultaneously obscure and reveal useful data.
They can help lead to serendipitous discovery through browsing, however, by an item appearing in one sub-menu and not another it can lead to useful data not being discovered if the archiver's mindset is different to that of the seeker.
Librarian's have worked on classification systems since their inception and be it Dewey, Library of Congress or any of the more specialist scheme they all have their issues. Someone researching a specific individual who has written poems, scientific literature and novels could have their work in very different locations in a Library or could all be grouped together depending on the scheme used.
As you can see the archiver's choice of hierarchy directly effects the efficiency that the data can be retrieved and the potential for serendipity.
Now while hierarchies of data can theoretically have plus points I feel that any time spent on the creation of hierarchies would be much better spent on creating superior meta data and enhancement to the search engine that is used to harvest from this data and cross information links between appropriately linked articles.
By enhancing metadata and improving search engines then my previous example proves this important point. A person researching the example author would instantly gain access to all their works and be made aware if they were not already that the author had a diverse catalogue of items. Anyone who was looking at poetry would be made aware that the author worked on other areas and may be curious to investigate them and how they may have influenced the work that they are investigating. A set hierarchy makes the only one of these outcomes possible, a good search and meta data leads both outcomes to become realistic.