[WEB4LIB] RE: website stats analysis

[Norwood, Randy]

I load my access logs into a MySQL database, partly for the reasons you
mention (more flexibility). I wrote a Perl script that parses and loads
the logs.

The questions you raise about the value of visitors, and how those are
determined, are somewhat complex, and are discussed in detail on various
websites (including webtrends/netiq). I think there is a lot of value in
the number of visits or "sessions". The log analysis tools such as
webtrends use various algorithms and assumptions to reconstruct the path
an user takes through the site during a session. These techniques would
be dfficult to replicate or improve on your own, and there wouldn't be
much point, since the analysis tools already do this about as well as it
can be done.

I can see what you're saying about summaries for visits obscuring the
internal vs external user ratio. You could prefilter your logs (using
Perl or SQL) into a public/staff log and everyone else, and then run
Webtrends on each indivudually. This assumes your campus network's IP
address blocks are assigned by building or some other small unit. If
everyone on campus uses DHCP and shares a common pool of IPs, the best
you might be able to do is on- vs off-campus.

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~
Randy Norwood
Web Manager
Texas Tech University Libraries
Office: 806-742-2238 x318
Fax: 806-742-8669
E-mail: randy.norwood at ttu.edu




-----Original Message-----
From: web4lib at webjunction.org
[mailto:web4lib at webjunction.org] On Behalf Of Jimm Wetherbee
Sent: Friday, August 06, 2004 1:36 PM
To: Multiple recipients of list
Subject: [WEB4LIB] RE: website stats analysis

Michele,

Hello, I'd like to hear from people who are responsible for analyzing
their website statistics.   

[->] Have never used a separate tool (though I sometime wish I had one)
bu have importing my logs into MS Access and using queries for the
analysis.

In using WebTrends, I find very little useful information other than the
information provided for "visitors." 

I can't, however, see the value in the numbers for "page views,"
"visits," and "hits," mostly because there is no way to eliminate staff
and public computers. And anyway, what is a "visit?" WebTrends says it's
all activity for one user.  So when "average visits per day" is 300,
what does that mean??? 


[->] I have a range of IP address that I know are local and run a query
that returns all local hits.  There are some limits to this.  Prior to
any analysis, I remove all entries aside from 50x and 20x for pages.
Since the number of local machine is restricted, they often hit the same
pages so the number of 304 entries (which I delete) would be
disproportionably high.  I loose some valid activity.  
[->] Still, what you could do is create a query that pulls out all local
hits and then exclude those hits within a query that would look at hits
for a given page or directory. 

---
[This E-mail scanned for viruses by Declude Virus]