The focus of this paper was federated search. Federated search works across multiple repositories – typically the extranet, EDRMS and intranet – and returns the results together.  The user interface presents the results either sliced by repository (top 3 intranet searches; top 3 EDRMS searches) or by relevance using tricky behind the scenes algorithms.

The key challenges faced by vendors developing search tools are:

·        De-duping (removing duplicate entries)

·        Developing algorithms that generate useful relevance rankings

·        Reflecting the security models present in each repository

When a tool meets these challenges then a vendor has a ‘viable’ federated search tool.  However, that doesn’t account for that nasty thing: human behaviour.  Letting humans at your repository can really stuff the accuracy of the search tool.  The nasty things that humans do include:

·        Changing content types

·        Adding and removing directories

·        Creating new workflows

·        Creating versions

·        Opting out of metadata

·        Changing taxonomies

All of these can conspire to seriously challenge your search tool.