Recording of Web Search Behavior

---

The pre-programmed navigation functions provided by Web browsers are well known and have been the subject of considerable IT/IS research including that of Cockburn and Jones (1996), Tauscher (1996), and Tauscher and Greenberg (1997). However, a more detailed taxonomy of search actions, information sources and the volume of information acquired is required to capture unconstrained Web search behavior. The variables are grouped as follows:

• Link variables: these describe the hypertext jumps taken by users from document to document in terms of the origin of the transition, that is, the point of departure, for example taking a link from a search engine or a Web site. The measurement of the number of links taken is a basic measure of hypertext activity.

• Site variables: these quantify the number of unique sites visited and the number of site visits made. This parallels previous survey research variables such as 'number of stores visited' (e.g., Bruce and Dommermuth, 1968; Katona and Mueller, 1955). The number of (different) sites visited is a measure of the breadth or scope of search. Conversely, the number of links taken within sites is a measure of the depth of search. The number of site visits is another measure of the 'breadth' of search, but this measure includes the accidental or deliberate repeat visits to sites that can occur during the information review and decision process.

• Search engine variables: search engines are a vital element of the Web environment. They also serve as landmarks and as information directories in addition to their search engine function. Search engine variables should record the production of search engine outputs separately for both categories of search engine. The number of times search engines are used, and the number of different search engines used by an individual may be readily identifiable characteristics of Web search behavior. Similarly, the refinement of a search engine query to reduce the number of Web sites included in the search engine output may also be a characteristic of an individual's search behavior.

• Navigation variables: navigation around the Web is a challenge for users and is of vital interest to researchers (Hoffman and Novak, 1996a). Web navigation requires the use of a Web browser. Web browsers uniformly provide a number of pre-programmed navigational functions. Navigation variables record all the Web browser based navigation actions of users during Web-wide information search. Favored hyperspace navigational techniques have been found to vary between individuals but may be a relatively fixed personal characteristic (Beishuizen et al., 1994). Similarly, an individual's methods of Web browser navigation appear to be a relatively fixed behavior unless effort is expended to increase their navigational repertoire (Tauscher and Greenberg, 1997). The recording of Web search actions, their number, and the search duration has the potential to be an informative method of studying individual Web search behavior.

In addition to the direct measurement of Web search actions, a number of other variables may be calculated and used to analyze search behavior. The derived variables include the mean number of search actions per minute, the percentage of search actions that yield data, and the percentage of search actions that yield new data. Apart from recording Web search behavior using the above taxonomy of search actions and information sources, it is also possible to characterize a Web search on the basis of its breadth, depth, speed, utilization of search engines and other variables. The characterization of search behavior and the investigation of search styles, especially in relation to individual behavioral traits, could be a fruitful area for research.

---