OR/MS Today - August 1997 - INFORMS Online

OR/MS Today, August 1997

Mystery Leads to Unexpected Benefits

By Michael Trick

Whenever a person accesses INFORMS Online (http://www.informs.org/), our web server creates log file entries documenting the server use. This information is not tremendously detailed, but the aggregate information says a lot about INFORMS Online and its users. To analyze the log file data, we use three programs: wwwstat (http://www.ics.uci.edu/pub/websoft/wwwstat/) and analog (http://www.statslab.cam.ac.uk/~sret1/analog/) are run daily, while wusage (http://www.boutell.com/wusage/) is run weekly. The results from all of these are available from the main page of INFORMS Online.

Rather than go through these extensive outputs in detail, let me tell you about a mystery identified by one program that lead to a solution found by another program which lead to a simple design change that radically changed how many people used IOL (and pointed out an unintended advantage of an earlier design decision).

The mystery arose almost from the first time the analog program was run. One number that analog gives is the number of distinct hosts served. A host is a machine name. For instance, I generally access the web through my machine "mat.gsia.cmu.edu," which is a single host. I sometimes use another machine, "kaos.gsia.cmu.edu," which gives another host. The server logs do not know user names, but they do give host names. In general, the number of host names is a rough estimate of the number of different people who access the server. It is only an estimate, since the same user may use two different machines (like my example above). Machines with multiple users, however, will only appear once. It seems reasonable to assume that these effects roughly cancel, giving a pretty good estimate.

The first time we ran analog, we were stunned by the very large number of machines that access IOL. For instance, in the three-month period beginning April 1, almost 100,000 different machines accessed IOL. Not bad for a 13,000-member society! Who were all these people? Why the interest in IOL? Of course, we were delighted that so many people found our corner of the Internet of interest, but it seemed strange to have so many different people!

The program wusage both identified the source of the interest, and a problem with the design of IOL. One statistic wusage provides is a list of "Most Common Trails" which tracks the most common ways a user (technically a host in a short period of time) traverses IOL. For instance, the most common path is always something like: "/, /Dir/Dir.html, /cgi-bin/SFgatesort" which means the user begins at the main page of IOL (/), moves on to the membership directory (/Dir/Dir.html) and then does a search (/cgi-bin/SFgatesort). Now, when we asked for the most common paths, there were hundreds of paths that looked something like "Conf/WA96/TALKS/WB09.2.html, /Conf/WA96/TALKS/WB09.html, /Conf/WA96/TALKS/WB09.2.html." This trail represents a traversal through a session at the Washington National Meeting. First a talk was examined (WB09.2), then the session information (WB09), then the talk was returned to.

Now, there were a number of interesting points: there were hundreds and hundreds of people each week accessing IOL who did not go through the main IOL page. Further examination of the logs showed that they were coming from search programs like Lycos (http://www.lycos.com/) and Infoseek (http://www.infoseek.com) directly into conference presentation pages after searching on terms of interest. When you consider that we now have more than 10,000 conference presentations indexed, it is not surprising that almost any search could lead to some INFORMS presentation. While we had known that people could do that, we were surprised by the number of people who were doing that.

Furthermore, when we looked at what IOL looked like for those coming in "through the back door," we were not particularly pleased. Many people were getting stuck, with insufficient navigational aids to allow them to visit the rest of IOL. We immediately made certain that no one can get stuck by including return pointers on all pages. A simple and obvious idea, of course, but we were rewarded immediately with a much higher count on people accessing the main IOL page, and wandering through the rest of INFORMS' offerings.

This mystery and its solution also points to an unexpected member benefit. When you present a paper at an INFORMS national conference, we make certain that your abstract is included in all the major search engines. There are many people out there using the Internet to find information, and many of them need the sort of information that we in OR/MS provide. IOL and its conference activities provide a link between prospective users and OR/MS professionals.

What's New on IOL

We have had a number of new associate editors join us. Welcome to John Quillinan in Member Services, Kent Sandoe in Aesthetics, Jay Aronson in Visibility and Placement, Steve Gould in Outreach and Gary Koehler in Technical Reports and Presentations.
A large number of organizational documents including the Internal Directory (http://www.informs.org/Intdir/contents.html), Committee Pages (http://www.informs.org/General/Comm), and INFORMS Department Pages (http://www.informs.org/General/Departments/departments.html).
The Student and Education Page (http://www.informs.org/Edu/edu.html) has been reorganized. Be sure to check it out for an announcement on the Student Union, an exciting activity of many INFORMS groups to provide a large collection of pointers and information of interest to students.

As always, comments, suggestions and criticisms can be sent to me at trick@cmu.edu

E-mail to the Editorial Department of OR/MS Today: orms@lionhrtpub.com

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-696
E-mail: lpi@lionhrtpub.com