Passive Observation of Log Data

This section shows results from a variety of Web sites that demonstrate the inferences one can make from merely observing the log data. Software products such as WebTrends <www.webtrrends.com> produce log file reports and one can work the raw log files as well.

Time of Day

Figure 11-4 shows average hourly accesses of two Web sites, by weekday vs. weekend . The site access patterns differ with the top graph being the communication site in Figure 11-2, and the bottom site being a content provider. The time axis is the time zone of the site itself. Figure 11-4 suggests that the two sites differ in access from work or home. It is also possible to see a dip around lunch time for both sites. Although neither of these sites is particularly busy, their graphs illustrate the possibility of collecting clean and beautiful data!

Figure 11-4: Hourly page views for a hospitality site (top) and for a content site (bottom)

Bookmarking

Bookmarking a site (adding the site to the Favorite's list in Microsoft terminology) implies a milestone in the relationship between that visitor and the site (Garofalakis et al., 2002; Murphy & Hofacker, 2003). When this event occurs, Microsoft's Internet Explorer browser requests the file favicon.ico, a 16 —16-pixel icon that the browser then uses for that site's entry in the visitor's favorites list. According to recent trade publications , Internet Explorer has 96% of the browser market (Murphy & Hofacker, 2003), suggesting that requests for favicon.ico indicate that visitors bookmarked the site.

This method works irrespective of whether the file favicon.ico is on the server. If the file is on the server, it creates a log entry when the server sends the file to the visitor. Otherwise, it creates a 404 File Not Found record if the favicon.ico file does not exist. In either case, the researcher has a valid marker that the user bookmarked the site.

How Did They Get Here?

The referrer URL shows if visitors arrived on the site through links on other Web pages or from a Web search service. If the latter, one can also determine the search terms used as these appear in the log file entry. A typical example might be:

192.18.209.138 - - [15/Nov/2003:10:31:29 -0500] 'GET / biking.html HTTP/1.0'

200 2444 'http://www.google.com/search?hl=en&q= fast+cyclists+in+ australia '

'Mozilla/4.0 (compatible; MSIE 6.0; AOL 8.0; Windows NT 5.1)'

In this case, a user entered the search term 'fast cyclists in Australia' into Google, and requested the Web site's 'biking.html' page.

No referring URL in the log entry also provides useful information. In addition to following links, visitors can type the URL directly into the browser, follow a previously made bookmark or have the page as the default opening page on their browser. The former, following links, suggests visitors searching for a page while the latter cases suggest familiar visitors.

A researcher could leverage referral information to assess the impact of off-line materials. For example, the firm could include a unique URL on the material, such as www.ourcompany.com/special. The log file would show visitors who entered the site by typing that URL.

Site Navigation

On what pages do visitors start, go and finish? After identifying individual visitors, one can follow their navigation through the site. For example, the hospitality site in Figures 11-2 and 11-3 discovered that visitors rarely started on the home page (Murphy et al., 2001). Most visitors arrived on a page of maps three levels into the site and then left. Management redesigned this map page and subsequently increased pages visited. This simple example exemplifies the myriad possibilities for using basic log file data to redesign a Web site or Web page to meet management goals, the topic that follows .