trouble sleeping

I had trouble sleeping, so I updated the buzz page to include links to transcripts.

This works because we are keeping the scrubbed transcripts (with no librarian or patron's personal information) in a local database. This might seem off track of me to spend time working on this, but it's not really too much trouble, it's fun, and it's important for a couple reasons.

First, I turned on the capturing-in-the-database script to help organize our evaluation of 24-7 refernece services from Tutor.com. We should all be looking at transcripts anyhows, and it will be good to eventually incorporate some librarian-to-librarian commenting through the buzz page.

Second, the pre-conference at VRD I attended today dealt, in part, with analyzing web site logs to understand user behavior. The idea is that we can look at the logs on our library websites to help us understand how our patrons use the library online and how they got there, including what search engines they use and the words they used to search. Jeff Rubin from Syracuse University recommended WebTrends, but there are many packages out there.

Ok, this is interesting, but I don't have a library website that people are coming to. I have a website for a statewide digital refernece service, and it is really most important to me that patrons visit L-net through partner libraries, not by searching Google. I also have something better than web statistics to find out what people are looking for, and that is our chat reference transcripts.

A quick query shows me that of 864 transcripts in that database (since mid October), 126 mention the word 'Google' somewhere in them. EBSCO, 32. Wikipedia, 25. Multnomah, 69.

This is a start, but what I think will be really helpful is finding out, for any user or population group, how many people are using co-browsing, how many are just using page-pushing, and how many are simply chatting? And I think we can do a lot of this with an automated process.

D.L. Cohen Associates' report from this summer also had some recommendations about evaluating questions by categorizing subjects, and I think analyzing the transcripts automatically can accomplish some of this.

The rest of the pre-conference was also good. I was more impressed with the speakers than all of the topics; I think that's something to aspire to.

There was a great comparison of vendor features by Jody Condit Fagan from James Madison University, and I will be sure to share the chart she made when I get ahold of it electronically. Jody also had some ideas about what questions we should be asking our vendors that didn't have to do with features, like can we have a copy of our data from their database (and now I really want it).

I also became convinced that when we look at vendors again this spring, we should look at not only what features we want, but what features we actually use, again aided by our transcripts database.