Analyzing OPAC logs

[Fernando Gómez]

Hello!

I am planning to start a detailed analysis of our logs, which have been
accumulating for more than six months since we launched a new Web PAC. The
purpose of this mail is to ask for some ideas about this task: guidelines,
pointers to interesting or innovative works in this area, or whatever you
would like to share. I thought it could be inspiring to know how others are
processing their own data (and if that processing leads to some kind of
improvement).

Our system has been designed in-house (using Bireme's WWWISIS technology),
so we have great freedom to decide what information to save in the logs. Up
to now, we are logging:

* date & time
* IP number (REMOTE_ADDR)
* query
* database (i.e. books, videos, periodicals, etc.)
* type of search (author, title, etc.)
* no. of hits
* no. of page solicited (we show 10 records per page)
* source of the query (an expression typed in the search form, a link
appearing in a previously displayed result, etc.)

This seems enough to harvest a great deal of useful information regarding
user behavior, failed searches, shortcomings in the system, ...

A concrete point that puzzles me a bit (not exactly library-related): what
concrete information can you extract form the list of IP numbers (the
REMOTE_ADDR variable)? Does the count of different IP numbers mean anything?
Is there any geographical information you can extract from them?

Well, thanks in advance for all your help. :)


Fernando Gómez
Bahía Blanca, Argentina