Tuesday, May 20, 2008

Enterprise Search - A Pragmatic View

These are my notes from a panel on enterprise search. One panelist pointed out that enterprise search may be reaching too far. They will not be able to get every silo. You should just try to get most of the silos.

The first question was "make it like Google." The first response was to have a simple interface that delivered relevant results. Relevancy is the hardest part. We do not have the magic special sauce that makes PageRank work so well for Google. One panelist also said that people are looking for Amazon.com. They want an ability to manipulate the search results to filter down to useful content. Another panelist showed the importance of putting the filters on the left-hand side. People are so use to Google putting advertisements on the right-hand side that they ignored the filters if they were on the right.

One issue is how to deliver relevancy or whether to include relevancy. What are the keys to relevancy.

What about the relationship between the document management system and the enterprise search? Most of the knowledge lives in the document management system and email. One panelist pointed out that those two sources are largely limited to work product.

What a bout taxonomy? Knowledge management has been focused on taxonomy. Does enterprise search reduce the need for a taxonomy? The panel all felt that taxonomy was very important to make enterprise search work. [I disagree. You want metadata to filter results. You do not necessarily need a rigid taxonomy.]

There was a lot of discussion about enterprise search inadvertently revealing documents that should not be public. One firm mentioned that they shut the enterprise search down after the a few weeks to give a cooling down period. During the cooling down period, people could put security on documents. Previously they had "security by obscurity." One firm had a policy to report information that is available through the search that should be secure. Another firm did not include administrative documents as part of the initial roll-out of the enterprise search.

The panel thought that any enterprise search is better than no enterprise search. They do produce different results and have some different capabilities. But they all seem to produce much better results than not having one. Some notes are focusing on multi-threading. You do not want people to have to wait in line for their search to run. Another point to focus on the index size. Some indexes are almost as big as the underlying content repositories.

What to include in the enterprise search? The panel responded:
  • Lotus Notes, resume collections, CRM, intranet, KM methodology system
  • Lotus Notes, Elite billing information, CRM, records, DMS, website
Each of the panelists each had the results tabbed out and not intermingling the results in a federated list. People mostly deferred to the default tab of the search results.

There was some challenge to the request for "The Google." Several audience members think the results need more bells and whistles to work. [I struggle with the question of whether the search tool needs to be more complicated or whether we need to store the information in a better way that simplifies the search.]

How do you prove value of enterprise search? One answer is that is an infrastructure thing, more than an ROI thing.

The panel offered their top advice on choosing and implementing and enterprise search tool:
  • Define the sources to search
  • Focus on security recognition
  • Understand who is searching for what, in what business context
  • Tune-ability of the tool
  • Focus on crafting the user interface
  • Put it into the toolbar of other applications; bring the ability to search to them

1 comment:

  1. Interesting post, Doug. We are currently undertaking a Proof of Concept for one of the leading enterprise search engines. If that goes well, we will proceed to a beta test and then launch, so I can let you know how it goes.

    Although, indeed, there is potentially valuable work product in the DMS, we hope to integrate other repositories (including Elite and library catalogue) to give users a more integrated view of relevant content.

    I do agree, however, with your comment that the ability to add rich metadata will ultimately be more useful than using taxonomies. In my previous firm, we found that the large majority of users used search rather than browse to find content, so the better the search tool, the happier the user.

    Going forward, I think people will become much more accustomed to adding tags and comments, which will add depth and context to the search results.


Note: Only a member of this blog may post a comment.