Eric Grose, Larry Varner and Chris Forsythe
emgrose@sandia.gov,
ljvarne@sandia.gov,
jcforsy@sandia.gov
Statistics and Human Factors, Sandia National Laboratories, Albuquerque, NM, USA
Recent estimates indicate that Intranets ¾ defined as private networks based on Internet and Web standards ¾ are being used by 59% of US and 38% of European organizations (IDC, 1997). In 1998 these numbers are projected to have grown to 77% for US and 75% for European organizations. As intranets permeate public and private sector organizations, substantial funds are being invested to develop web-based products for disseminating information and conducting internal business. These products fall into two categories: content-based web pages that present information and feature minimal interaction. Examples of content-based pages include news, manuals, presentations, catalogs, FAQs, etc. The second category of web products are interactive applications such as timecards, order/request forms, employee/service directories, etc. Utilizing the corporate Intranet at Sandia National Laboratories (SNL), it has been observed that of several hundred thousand accessible pages, the vast majority of Intranet traffic targets less than one percent of them (Fig. 1). This figure dramatically shows that while some web pages are wildly popular, many do not justify the cost to build them.

Figure 1. Number of Requests in a Month for the Top 100 Most Accessed Sites at SNL
This paper attempts to explain why some products are accessed so often and others are not by presenting a model of factors that contribute to a web product’s "penetration" into a user population. Such a model would be extremely useful because it could be used to predict a product’s potential for success before resources are committed. As the above figure apparently shows, the ability to make such predictions is greatly needed.
This paper reports the initial phase of model development. Parameters thought to affect penetration are presented and, where possible, forensic evidence from Sandia’s intranet is evaluated to justify the parameter’s inclusion in the model. These evaluations are based on usage patterns of the SNL Intranet over the past three and a half years. The model incorporates thirteen parameters (see Table 1). No priority is implied by the order of their presentation. Future evaluation may suggest additional parameters or higher-level factors that justify consolidation.
|
Relevance |
Entertainment |
|
Forced Compliance |
Awareness |
|
Tactical Advantage |
Habit |
|
Productivity Improvement |
Competition |
|
Mistrust |
Novelty |
|
Compatibility |
Exploitive or degrading work |
|
Social Connectedness |
Table 1. A Model of Penetration
Relevance
The penetration of a product will be partly determined by the proportion within a population for whom the product is relevant. All other parameters being equal, a product that is relevant for every member of a population will exhibit substantially greater penetration than a product that is relevant only for a subpopulation. This parameter is illustrated through a comparison of Newsletters accessible through the SNL Intranet. Figure 2 shows the level of usage for nine corporate Newsletters. All are distributed only through the corporate Intranet and none have a paper analog. The two with greatest usage report a variety of news items, some of which are relevant to all lab employees. The remaining Newsletters report items relevant to specific organizations or job functions.

Figure 2. Requests for Nine Corporate Newsletters as a Function of the Size
of the Relevant Population.
Forced Compliance
Within a corporate environment, attempts to ensure penetration may entail various mechanisms by which users are required to use a specific product. SNL contracted with Netscape Communications Corporation to develop a version of the Netscape Navigator 3.0™ browser with specialized security features and an interface customized for the SNL Intranet. This browser was adopted as a corporate standard and when SNL introduced electronic timekeeping, the timecard application would only run on the corporate standard browser. Employees were required to complete their own weekly timecard, meaning that each employee had to use the corporate standard browser. In this case, it might be predicted that nearly 100% of the browsers used at SNL would be the SNL version of Netscape Navigator 3.0™. However, it was discovered that approximately 12% of accesses to the SNL Intranet involved non-standard browsers (e.g., Netscape Communicator 4.0™, Internet Explorer™). This finding suggests an upper bound to the level of penetration that can be achieved through forced compliance. This is consistent with literature suggesting that despite imposed standards, forced compliance alone is insufficient to obtain complete penetration (Bikson & Eveland, 1990).
Tactical Advantage
Sandia’s Weather page is one of the most frequently visited web sites. One of the reasons this is so, we propose, is that employees check the weather forecast in order to plan around bad weather. That employees check the Weather page for their own tactical advantage seems supported by Figure 3, which shows that the lowest average temperatures corresponded with the greatest activity on the Weather page.

Figure 3. Requests for the SNL Weather Page Compared to Average Monthly Temperature
Productivity Improvement
At its simplest level, this parameter refers to individual productivity gains. On the SNL Intranet the most frequently accessed pages are two employee directories which search out an employee’s phone number, location, etc. They differ in that one, "Quick Phone", allows just the employee’s last name to be entered while the other, "Sandia Directory", allows a person to be searched by more criteria (department, building, etc.). These simple applications receive three times the traffic of the next most frequently accessed URL (SNL Daily News). A simple comparison revealed that when searching for an employee using Quick Phone there was a one-third time savings using the online directory, as compared to a paper directory. This benefit is amplified by the knowledge that the information available via the web is likely to be current, whereas the paper directory is often out-of-date. Furthermore, when conducting a search by organization, the web-based directory yielded an almost 100% time savings. Given that the two directory applications are used several million times a year, it is ironic that they were developed for marginal costs.
Another tool that has penetrated deeply is e-mail. Figure 4 shows that as e-mail has penetrated within Sandia it has displaced a more cumbersome method: hardcopy memos that were routed through Sandia’s mail room.

Figure 4. Increase in E-mail Compared to Conventional Mail Over the Same Period
Mistrust
The reluctance of users to transmit credit card information via the Internet testifies to the importance of trust. If a transaction entails sensitive information (e.g., credit card number) or there exists the potential for highly undesirable consequence (e.g., time-sensitive monetary transactions), penetration will be strongly influenced by factors associated with trust and confidence. In these conditions, penetration will be more likely with Web products that provide a sense of assurance regarding security and/or reliability.
Compatibility
For work that requires the electronic sharing of information, productivity gains seen at a local/individual level may be outweighed by the cost of hardware or software incompatibility. For example, a desktop publishing application that provides valuable features for one user may produce files that cannot be easily read by co-workers with whom that user must collaborate. The effects of compatibility on productivity partly explain the phenomenon that superior technology often does not penetrate.
Social Connectedness
Many sites fulfill a user’s need to feel part of a larger community. This may occur through news sites that promote common knowledge of community members and an awareness of community events. It may also occur through sites that enable dialogue between community members.
Entertainment
Although a theory of what constitutes entertainment is elusive (Kanerva, et al, 1998), the high access rates for WWW sites such as Dilbert and ESPN SportZone illustrate that entertainment may produce penetration.
Awareness
Awareness of a product may come through various channels. We are all familiar with mass media advertisements as a means of attaining broad-based product awareness. Word of mouth is another mechanism, and one that contributed heavily to the initial penetration of the NCSA Mosaic web browser. Broad awareness of a content site or application should generally lead to greater penetration. A demonstration of this principle comes from the effect that advertising on the SNL home page had on a particular Web product, "Favorites". This web page contains links to the most frequently trafficked sites on Sandia’s Intranet. As Figure 5 shows, requests for this page began to level off until an icon link (a "teaser") advertising the page was placed on the SNL home page, after which requests for the page rose dramatically. The teaser was removed after two weeks.

Figure 5. The Effect of Awareness on Penetration of a Web Page
Habit
Nearly all innovations must unseat a predecessor. If users are loyal to existing technology, penetration may be delayed and diminished, despite a new product’s superior performance. This principle is illustrated by the previously mentioned SNL directories. The application built first, Sandia Directory, offered the ability to search for someone by several criteria. It was observed that users typically entered a last name and pressed the <return> key. The newer application, Quick Phone, caters to this method by offering only a search-by-name field. Both applications are available side-by-side on the corporate home page. A survey of ten users revealed that 60% choose the full directory, even though they only use the last name field and it is less efficient for this method of search. Users continue with the application with which they were first successful, the Sandia Directory.
Competition
As the total volume of web resources increases, the penetration for a given resource is expected to decline. This phenomenon is similar to the reduction in market share being experienced by the major TV networks as competition increases with numerous other outlets for news and entertainment. Figure 6 shows that the number of requests to SNL’s most requested pages, the Directories, steadily increased over a two-year period, even though their percentage of total requests to all resources on the Intranet decreased. This suggests that competition with other web resources is at work.

Figure 6. Requests to SNL’s Directories and Their Percent of Total Intranet Requests
Novelty
As a parameter in the model, novelty serves a complex purpose. It can explain why a particular product may first be used and also account for continued access of a product that does not currently contain information relevant to a user. Novelty in a product may serve to attract the attention of a user examining the functionality of a product. The novelty could be a result of a unique aspect of the user interface, unexpected functionality, or some other attention-grabbing technique. From this perspective novelty serves to draw the user into the product.
Novelty can also be used to induce a user to return to a product even though the current information available is not personally relevant. The possibility that the information available through a product will change and thereby become personally relevant may be enough to promote a return to the site. The novelty aspect lies in the ability of the product to provide new, current, and changing information to the user. The knowledge of novelty in a product requires the user to return to the site to view any changes that have occurred since the last access. This is true of products such as CNN, which changes its headlines every so often throughout the day, and PointCast™, a push technology product that involves constantly changing information that can be passively monitored or updated manually. In either case both products involve novelty that promotes return to the information to keep updated.
One way to demonstrate how novelty may affect penetration is to show that when the information on a page becomes stale, people stop coming back. Figure 7 shows the requests for a Computer Support newsletter that ceased updating in January ’96. Requests for this site tapered off almost immediately. (It’s notable, actually, how slowly this phenomenon transpired. In fact, the company timecard was introduced in October ’96 and it is our hypothesis that the requests over that period are explained by employees seeking help on this new application.)

Figure 7. Requests for a Newsletter After it Ceased Updating
Exploitative or Degrading Work
Technology acceptance is less likely under conditions in which workers perceive the technology as exploitative or degrading (Carlopio, 1988). With information systems and the automation of business practices, this may occur with applications that, in eliminating job functions, transfer a subset of those job functions to staff who would not ordinarily perform such activities. For example, positive user attitudes toward an application, and penetration, would be lessened if employees in technical professions perceive that they are being forced to perform clerical work.
Conclusion
The Web is transforming the working world. Intranets have provided a natural solution to the perplexing and expensive problem of providing an electronic infrastructure on top of incompatible hardware and software systems. The relative ease of producing web pages has also encouraged more people than ever before to contribute to the growth and variety of that infrastructure. Usability analysts need tools to evaluate intranet products to predict and influence their potential for success.
In this paper we have attempted to articulate factors that contribute to the success of a Web product in gaining acceptance among — to penetrate — a user population. Obviously there is much more to be done with the model. The next phase will be to develop procedures for quantifying model parameters. Best fit approaches will then be employed to ascertain the relationship between the parameters. Finally, validation exercises will emphasize using the model to predict relative usage for new web products as they are introduced to the SNL Intranet.
We anticipate that Intranets built with consideration of the factors discussed in this paper would look very different from the one presented in Figure 1, where only a few Web pages account for a disproportionate share of the overall traffic. Rather than accessing just a few Web products, users would have many superior products to choose from. This would have the effect of reducing the traffic to any one page and spreading it over many (see Figure 8).

Figure 8. A Possible Effect of Using the Penetration Model on Intranet Products
References
The Persuasive Intranet, IDC, Oct, 1997.
Carlopio, J. (1988). A history of social psychological reactions to new technology. Journal of Occupational Psychology, 61, 67-72.
Bikson T. K. & Eveland, J. D. (1990). The interplay of work group structures and computer support. In J. Galegher, R. E. Kraut & C. Egido (Eds.), Intellectual Teamwork. Hillsdale NJ: Lawrence Erlbaum Associates.
Kanerva, A., Keeker, K., Risden, K., Schuh, E., & Czerwinski, M. (1998). Web usability research at Microsoft Corporation. In C. Forsythe, E. Grose, & J. Ratner (Eds.), Human Factors in Web Development. Mahwah, NJ: Lawrence Erlbaum Associates.