Understanding Web Log Analysis
Sep 23rd, 2004 by Karen
Understanding Web Log Analysis
I just found a good article
on web log analysis from in a 1998 issue Network Notes a publication of
the National Library of Canada. Although old this article is still
relevant. It does an excellent job of describing the information in a
log file and what can be learned or inferred from examining log files.
The authors also talk about what information cannot be gathered from
log files, what information in incomplete in log files, what inferences
are dangerous to make from log analysis.
One inference that the authors found dangerous is counting sessions. They state
calculated by some log analyzer products by tracking requests received
from an IP address until a period of inactivity (say 30 minutes)
indicates to the software that the “session” has ended. As this
calculation is based on two unsound assumptions–that a host
corresponds to an individual, and that the individual would not
normally pause (whether to go to another site or another task) within a
site visit–user sessions are, at best, gross estimates.
While there is some truth to this statement, user
sessions can be used to make comparisons to see if overall web site use
or the use of an individual page is increasing or decreasing over time.
Sessions can also show differences between use of individual pages.
However, it must be remembered that this can't be used as “exact”
numbers as they really are estimates. So this information is not useful in all instances.
Check out this article, it is well worth your time if you are trying to get an handle on web log file analysis.
