2005 Poster Session

Web Log Analysis and Resulting Improvements

David Stern, Kline Science Library, Yale University
Better Understanding Your Users (Poster Session)
cosponsored by the Chemistry, Sci-Tech, and PAM Divisions
SLA Annual Conference, Toronto, June 6, 2005

Abstract: Many organizations already maintain web transaction logs as part of their basic web server software. Other transaction log software products provide more sophisticated options in order to identify more detailed user patterns. This poster session displays selected user data screens and discusses modifications that were and might be made based upon such analysis.

BASIC DATA

Basic use data includes:

type/version of browser used,
country of origin,
visitor user domain (edu/com/robot, etc),
actual pages visited (in some cases, from which page you were referred),
hour of visit,
day of the week of visit, and more.

This basic use data is worth reviewing. Simple user assessment can be performed and modifications to web sites and related services can be considered based upon this data.

ADVANCED STATISTICS ANALYSIS

In addition to simple hit analysis, more advanced analysis can include the following: search phrase review, Top Paths through the site (page-to-page user trails), observing entries by referring page/service, and comparing top hits by individual dates and/or in designated (e.g. weekly, monthly, semester) aggregates.

Top Paths statistics page

In one example, we observed what must have been repeated user frustrations due to initial visits to a "dead-end" information-only page for a proprietary product (CrossFire). We re-routed first time users to the download page and then observed additional requests for password access and download assistance.

In another instance, we observed increased use of appropriate chemistry web resource pages (and elevated long-term use) immediately after course-integrated classroom instruction. We may soon modify our subject pages, adding specific tools to the subject Help sections based upon the time of semester.

COUNTING LINK USAGE ON A PAGE:

There is also the possibility of reviewing the use of individual links on a particular page using counting scripts. Yale Libraries analysis showed virtually no use of the highly touted "Quick Links" on our Home Page templates ... raising questions about more effective utilization of such high visibility real estate.

Single Page Link Use Analysis Page

REVIEWING OPAC USAGE:

We were able to review the type of clients that are used to search our OPAC and the quality of searches that are performed. As a result:

We modified our Keyword searches to include an implied "and" which resulted in less precise but much more effective searching.
We were also able to discover a very large number of Z39.50 searches. These might be explained by "outside" library searches, EndNote connection searches, and users preferring telnet quick results and downloading options. This has begun a complete reassessment of our support for non-Web interface support.

OPAC analysis page

REVIEWING SEARCHES

Making search option modification

We determined that many users were incorrectly using the "library web search" tool to search for WWW searches. Despite extensive verbal warnings that this was inappropriate, people ignored the messages and went straight to the search field.

In addition, our use of the "Google" icon to show we use their search engine made it even more confusing.

Old "Library Web Site Search" Page

We modified the Search screen, removing the Google icon, and requiring the selection of a specific search domain. This allowed for greater options that were appropriate for a wider variety of search types.

New Multi-domain Search Page

David Stern
Director of Science Libraries and Information Services
Kline Science Library
219 Prospect Street
P.O. Box 208111
New Haven, CT 06520-8111
phone: 203 432-3447
fax: 203 432-3441
email: david.e.stern@yale.edu

date of last revision 7/5/05

Comments

Page updated: