Time out for a video interview (which will appear on the FASTforward blog sometime today or tomorrow), then back to the business productivity track to hear Auli Ellä of Orion, a Finnish pharmaceutical, discuss their enterprise search implementation. As a research-based company, internal users need to be able to find information both internally – in a Documentum document management system, eRooms and the intranet – and externally.
As she put it, it’s not about searching, it’s about finding. Prior to implementing FAST, users didn’t know which system to look in or what metadata to use, and failed searches (where they didn’t find what they needed) took an average of an hour each, which tells you how them cost-justified their enterprise search project.
They started with a pilot enterprise search project that used a small selection of internal and external information sources accessed by a select group of R&D and business users, and used that to prove their business case and gain management acceptance.
She outlined two major benefits that they’re seeing from enterprise search: it supports decision-making, and supports innovation. It saves them a lot of time spent searching internal and external resources, but also adds value to search results through increased accuracy and reliability, and the ability to filter, categorize and drill-down into the results. Information is re-used in place, not copied between systems, using the existing metadata from the source systems both for indexing and for filtering results. This provided an easier and faster search that is at least as reliable as a single source search, and although search results are federated, the access rights of the original information source are respected such that a user can’t see search results for content that they would not have access to in the source system. Users could create and save searches, making it even faster to locate frequently-accessed information.
The user feedback was impressive: in fact, some users thought that it couldn’t be real because it was too fast.
She had some lessons learned for implementing enterprise search:
- It’s essential to start with a pilot, doing a concrete proof of concept, then implementing iteratively
- It’s important to have a diverse group of pilot users, not just highly-skilled knowledge workers
- Keep it simple with the filters and navigation: too much sophisticated can just confuse things
- Be prepared for resistance, and combat that by offering tips and tricks for using the system more effectively
Going back to the goal of supporting decision-making, it’s helping them by allowing users to see all relevant information available on one screen when, for example, gathering materials for a meeting. On the innovation side, they’re starting to have some unexpected results – which drive innovation – becoming visible more easily by using filters such as content authors.
They’re now adding more internal and external information sources, and improving functionality based on feedback from the users for things such as selecting a single information source when the user does know where the content is – which is an interesting request, when you consider that that means that the users would rather search in FAST than in the underlying system due to speed, ease of use and functionality. Reliability and speed are critical for search, of course: without both of those, the users will reject it.
In the future, they’ll be developing new filters and search profiles, adding structured content, and incorporating search behind the scenes in portals to display relevant information (like the company cafeteria menu, which is currently the top search in their system). As this happens, however, they are aware that this continuous development can make it easier for the user, but increases the amount of work in the background by the information management people.
Her conclusions: if done well, search can become a personalized desktop where discovering, re-using and refining information is business as usual. It’s all about turning information into knowledge and action.