CHI 97 Electronic Publications: Late-Breaking/Short Talks
CHI 97 Prev CHI 97 Electronic Publications: Late-Breaking/Short Talks Next

How People Use WWW Bookmarks

David Abrams
Dynamic Graphics Project
University of Toronto
Toronto, Ontario M5S 1A4, Canada

Ron Baecker
Dynamic Graphics Project
University of Toronto
Toronto, Ontario M5S 1A4, Canada


In this detailed empirical study of WWW browsing and bookmarks we define a personal information space as having five basic properties paralleling those of a larger complex information space. We describe user behavior on the Web and show how a user's bookmark archive is a personal Web information space.


WWW, bookmark, information space, user study, survey, empirical study

© 1997 Copyright on this material is held by the authors.


Although WWW bookmarks, hotlists, favorites, or personal URLs are used by over 84% of Web users [2], they routinely have difficulty using bookmarks. This detailed user study of WWW browsing and bookmarks [1] identifies the fundamental principles underlying a complex information space and applies them to a personal Web information space.


We performed an in-depth analysis of how users mitigate the problems of the WWW through bookmarks. Our study began with informal interviews and a survey of 12 users. Next, we conducted a formal survey of 322 users at the Internet Beyond the Year 2000 conference. We collected and analyzed the bookmark files of 56 users, and analyzed the usage data of 23 users over a 6 week period (data from [3]). Finally, we surveyed 27 users in a mental maps study of a user's personal view of the Web.


Users find localities of interest, create bookmarks to return to them, and conceptualize their use of bookmarks through metaphors. This behavior leads to the development of a personal information space based on the bookmarks collected.

Users Find Localities of Interest

Users revisit groups of related Web pages and there is a 58% probability that the next page visited is one the user has already encountered [3]. For example, a variety of local navigation patterns will emerge while a user visits Web pages in a cluster of relevant documents [3]. We define a Web locality as a group of pages having a thematic organization, structural cues, and links representing explicit semantic relations.

Users Establish Bookmarks to Distinct Web Localities

Bookmarks are used in only 2.7% of all navigation actions [3], yet they are "essential" to users. Bookmarks are primarily used to jump between Web localities, then users employ a variety of methods to navigate within the new locality. Our analysis of user bookmark files found that 87% of bookmarks point to distinct domain names, independent of the size of the bookmark archive. In addition, bookmarks serve as launching points for hypertext exploration and local navigation patterns emerge after using a bookmark.

Users Conceptualize WWW Access Through Metaphors

Users conceptualize the use of bookmarks and browsing with four metaphors. First, users mark/tag Web pages by actively assigning distinctive labels through bookmarks (identification). Second, users pull Web pages out of hypertext into their collection of bookmarks (collection). Third, users travel through cyberspace and return to sites via bookmarks (movement). In the fourth metaphor, users recall previous episodes associated with pages they have stored in their bookmark archives (episodes).

Users Build an Archive from Bookmarks

Users create a personal archive out of the bookmarks they have collected. In the ideal environment, users cost-tune their environment so that frequently used items are easily accessible, while less used items are stored further away. A time series analysis showed that bookmarks are created and often not used for months.


Users of a complex information space like the WWW are (1) overloaded with information. This medium is (2) polluted with redundant, erroneous and low quality information. It progresses toward disorder according to the principle of (3) entropy. It has (4) no aggregate structure which organizes distinct Web localities. Users have (5) no global view of the entire WWW from which to forage for relevant pages.

Users employ a personal information space to counteract these five problems of a complex information space. (1) They prevent information overload by incrementally building a small archive. (2) They avoid pollution by selecting only useful items and creating a known source of high value. (3) They reduce entropy through maintenance and they organize only when necessary. (4) They add structure by cost-tuning their information environment. (5) They compensate for the lack of a global view by creating their own personal view.


A Web user's bookmark archive is a personal information space with five basic properties each of which addresses a unique problem indigenous to a complex information space.

Users Start Small and Build Incrementally (overload)

Users gradually buildup a small sized archive where 68% of users have 11 to 100 bookmarks. Bookmarks are added in small clusters and over 93% of users create 0 to 5 bookmarks each browsing session. The growth rate of a user's bookmark archive is linear (r = 0.9958, P < 0.001) over the first year and can be described by:

where N is the number of bookmarks in the archive and T is the number of days since the first bookmark was created.

Users Select Only Useful Items (pollution)

A bookmark archive serves as a known source of valuable Web sites because users select items according to five criteria: general usefulness, quality, personal interest, frequency of use, and potential future use. Over 95% of users with 20 or more items have at least one search engine in their archive. When bookmarking a query engine (e.g. Lycos) users bookmark the page for entering queries. In contrast, they tend to bookmark selected parts of the hierarchy when using a meta-index (e.g. Yahoo). They rarely bookmark the search results page.

We generated an extensive taxonomy of bookmark use based on user comments. Bookmarks enable users to avoid managing URLs, are used as a "mnemonic device" and an inter-session history mechanism. Users collect groups of related bookmarks in order to author a Web page out of those URLs and give presentations.

Users Add Value Through Organization (entropy)

Users expend effort to manage their personal information space, but tend to organize only when necessary. 99% of users with less than 35 bookmarks have 0 folders. Once the user's archive grows beyond 35, the approximate number of items that we currently can see in a pull-down menu, the number of folders correlates directly with the number of bookmarks (r2= 0.60),

which is significant (F1,26 = 38.52, P <0.0001). A larger archive requires more organization and the poor scalability of current tools force users to find new ways to organize bookmarks. Users with more bookmarks employ more sophisticated means of organizing. We identified five primary organizational methods: 8% manually re-arrange a list, 29% create a set of folders, 23% generate a hierarchy of folders within folders, 3% off-load to an external database or bookmark management program, and 2% author Web pages. 37% of users do not organize their bookmarks.

Sporadic-filers, 46% of respondents, schedule a special clean-up session to organize their bookmarks. End-of-session filers, 6%, set aside time after each session to organize. In contrast, creation-time filers, 21%, store each bookmark in its place just after interpreting its contents.

Users Structure for Retrieval (no aggregate structure)

The structure of a user's personal archive facilitates efficient retrieval based on expected use. Users encounter significant problems while managing a semantic hierarchy, naming folders, maintaining a stable ontology, and retrieving bookmarks from the semantic hierarchy (e.g. "what folder did I put it in?"). Representation issues hinder retrieval because the titles of bookmarks are often not descriptive.

Users Establish a Personal View (no global-view)

We asked a group of experienced users to each draw a mental map of the WWW. Abstract landmarks were prevalent in 70.4% of the 27 drawings collected from a sample of 72 people. 51.9% of the mental maps included a reference to search which suggests that many users' experience with the Web is mediated through search engines. Users separated "My Bookmarks" from the "cloud of unmapped resources" which appeared in many drawings. Users develop a personal, gestalt view of the Web through use.


A user's bookmark archive is a personal information space with five principles paralleling those of a complex information space. As a user's bookmark archive grows, it can begin to exhibit the properties of a complex information space. Users are continually challenged with maintaining an effective personal Web information space. Our future research will apply these results to the design of more usable Web browsers and bookmark management tools.


Special thanks to Mark Chignell for many provocative discussions about this research. Linda Tauscher provided us usage data and Bruce Homer helped with the analysis.


1. Abrams, D. Human Factors of Personal Web Information Spaces. MSC Thesis, University of Toronto, 1997.

2. Pitkow, J. GVU's 5th WWW User Survey, 1996.

3. Tauscher, L. and Greenberg, S. Revisitation patterns in world wide web navigation, in Proceedings of CHI '97 (Altanta GA, March 1997), ACM Press.

CHI 97 Prev CHI 97 Electronic Publications: Late-Breaking/Short Talks Next

CHI 97 Electronic Publications: Late-Breaking/Short Talks