Page 343 -

12.2 Existing tools 333

Interfaces (APIs), often following the Representational State Transfer (REST) con-
ventions (Fielding and Taylor, 2002). Essentially, REST (and similar) APIs define
structured web requests to return data suitable for extraction and manipulation by
third-party web sites and other programs. As these APIs can be used to provide data
to mobile apps and stand-alone desktop programs—as well as web pages—any such
accesses will be included in web logs, and can therefore be analyzed to track usage
patterns.
12.2.1.2 Web usability/design research
By telling us which pages were accessed and when, web access logs can provide
valuable information for usability evaluations and understanding of usage patterns.
Relatively simple page access counts tell us which pages are accessed frequently,
and which are not. When coupled with an understanding of page layout informa-
tion, this can help identify opportunities for improving usability. Aggregate counts of
timestamps, referrers, and user agents can be used to understand when a site is being
used, how people are getting to links within a site (external referrers are particularly
interesting in this regard), and which browsers they are using—all potentially useful
information in the context of evaluating a site design. Interactive visualizations of
this data at multiple granularities—particularly when coordinated with views of the
site—can provide guidance for improving site design (Hochheiser and Shneiderman,
2001). For example, if important areas of the site are infrequently accessed, links
might be moved to more prominent locations or be made more visually distinctive.
Postmodification analysis can be used to evaluate the success (or lack thereof) of
such measures.
Web access logs also provide the intriguing possibility of extracting information
about the actions of specific users as they navigate a website. This information can
be very useful for understanding which path users take through a site and where they
might run into problems.
As each entry in an access log can contain an Internet address, a timestamp, the
requested URL, the referring URL, and a user agent, we might be tempted to com-
bine this information with knowledge of a site's layout to infer the path of specific
users through a site. If we see that an access to “index.html” is soon followed by a
request for “help.html,” with both requests originating from the same network ad-
dress, we might think that these requests came from the same user. Matching user
agents and an entry indicating that the referrer page for the “index.html” page was
the “help.html” page might increase our confidence in this theory. Judiciously used
web cookies can provide additional useful information.
Unfortunately, things are not necessarily that simple. Firewalls and other network
address schemes may make requests that come from multiple users appear as if they
all come from the same machine. Web browsers can easily be configured to provide
misleading information for the user agent fields and referrer fields. Web redirects
may create misleading requests, appearing as if a user intended to visit a site, when
they had no such interest. Cookies may be disabled by some users and browsers.
Despite the problems, access logs can be used to generate useful models of user
paths (Pirolli and Pitkow, 1999). Augmenting these records with additional information