Maxine Armitage of Find It Information Management recently analyzed two years' of search queries from a government agency's intranet. She published her findings in a brief case study (80Kb PDF).

We were especially impressed by how Maxine normalized the queries, and asked her how she did it. Here are the steps, in her own words:

1. Alphabetized search terms together with their number of occurrences (information gathered from search logs).


2. Grouped obvious terms under the one topic, e.g., "flex time," "flex agreement," "flexi sheets" under "flex time".


3. Brought together terms such as "infringements" and "‘notices" under the more frequently used "penalties".


4. Read the list for abbreviations, jargon and grouped with topics arising out of #s 2 and 3 if applicable.


5. Used some intuition and knowledge of the organization to tease out other topics.


6. Was guided by what I was seeing; e.g., I noticed that this memo and that memo, and this circular and that circular were being sought. That indicated that I shouldn’t just restrict myself to topics, but should also look at "types".

That's good, practical advice to consider while analyzing search queries. And it's not rocket science, just common sense. Don't worry if you're clustering queries the right way or the wrong way: just do it and see what patterns emerge.

—Lou Rosenfeld