Analyzing search result counts
- Hi all - I have a question about what kind of analysis you might have
tried with data about the # of search results user queries return (if
We have the ability with our search solution to know how many results
were available for each query executed by a user. The most
significant (to me) are those searches that return 0 results - you
want to understand what patterns are in those searches that might
identify missing content or mistagged content, etc.
I've been asked for an analysis of the overall data for search result
counts for searches. As you might know if you read my blog (at least
when I was publishing regularly - which I hope to start again soon!),
you'll know that my own bias is that if you're going to bother
analyzing data, you should do so with the expectation that you would
act on it and that you would expect to be able (and want to) influence
what you're measuring. (Otherwise, what's the point in the analysis.)
I've done some preliminary review of the data we have for this metric
and I'm a bit stuck in thinking about what actions the insight might
Some examples: I can determine the average number of results a user
sees. I can determine the median number of results a user sees. I
can figure out a standard deviation of the numbers. I can chart out
the data to see if I can find patterns (and I can).
At the end of the day, though, other than the special case of 0
results, what sorts of things might be driven by understanding this
Let's say your goal was to reduce the median (or the average). One
way to do that is to reduce content in your search corpus. Is that
useful? Another way would be to look at it as more of a training
issue - you just need to convince your users to use more precise
searches, right (and good luck doing that :-) )!
Anyway - I'm hoping someone out there has considered this and has some
insights to share.
Thanks in advance!