An Ecological Model of Web Usage
Chris Forsythe, jcforsy@sandia.gov
Sandia Labs
Two years ago, Eric Grose and I began a series of studies to identify and model the factors that predict usage of web products. We had both worked on the design, development and testing of a variety of web products destined for a corporate intranet. These products spanned a range of job-related applications. Many provided web versions of content or functions that were well established prior to there being a corporate intranet (e.g., phone directory, newsletters, timecard). Others enabled functions that had no true predecessor (e.g., web-based configuration management systems). But in almost every case, development of the product was motivated by a belief that the product would return benefits well exceeding the costs associated with its development and subsequent maintenance.
Generally, with web products, whether destined for the World Wide Web or a corporate intranet(s), there has been a broad sentiment that, "if I build it, they will come." This may be partly a product of egotism. People commonly overestimate the degree to which others share their beliefs and interests. Also, there is often a presumption that there exists an abundant population of individuals who have ample spare time for yet another new web product. The web offers the impetus for countless ideas and has an unusual capacity to spawn optimism. Consequently, there is phenomenal growth in web sites both with the World Wide Web, as well as corporate intranets.
The reality for web developers is that a very small percentage of web sites experience heavy traffic, while the overwhelming majority languish, unnoticed and ignored. In previous papers and presentations, we have discussed the phenomenon illustrated in Figure 1 referred to as the "Hockey Stick" (Grose & Forsythe, 1998). Briefly, if web products are ranked from greatest to least usage, a small number of web products will exhibit extremely high levels of traffic with there being a sharp drop-off and the remaining web products receiving substantially less traffic. We have also noted that for a sample of web products, there was no correlation between usage and estimated development costs. In fact, some of the most frequently accessed web products are extremely simple database queries that required only a few hours to develop (e.g., phone directory, vacation balance query).
Figure 1. Traffic observed with the 100 most frequented sites on a corporate intranet.
Admittedly, usage and value do not have a one-to-one correspondence. It may be argued that a page out of a manual that requires fifteen minutes to covert to HTML format, upload and link is cost-justified if one person saves twenty minutes searching for the equivalent hard copy version. However, in reality, twenty minutes could be just as easily lost searching for the online version. Nonetheless, any business committing precious overhead funds to development of internal web products should be concerned about targeting products that will generate substantial usage. For this reason, we saw tremendous value in developing a model that would predict usage prior to product development. We believed that if correctly implemented, such a model would allow the "Hockey Stick" pattern of usage shown in Figure 1 to be transformed into a more gradual distribution similar to Figure 2.
Figure 2. Desired pattern of usage relative to actual usage.
Our efforts have been reasonably successful. As discussed elsewhere, we identified a collection of factors as candidate predictors and assessed them using known web products and available usage data (Grose & Forsythe, 1998). In the initial analysis, we predicted 87% of the variance in usage with three factors (relevant population, habit and competition) and their interactions. A follow-up study sampled a different collection of web products and accounted for 67% of the variance in usage with the same three factors and their interactions. It is in retrospect that we have begun to question, not our results, but our general conceptualization of web usage. In particular, we believe a model that emphasizes the parallels between ecological phenomena and patterns of usage observed with web products may offer insights that were absent in our earlier conceptualizations.
Ecological Model of Web Usage
Ecological systems contain many individual living organisms which on the basis of genetic similarities, may be grouped into distinct species. The Internet, or an intranet, is populated by web products that are comparable to individual living organisms. These products may be roughly grouped into types, or genre, such as news sites, chat rooms, e-businesses, etc. However, the lines of distinction between types of web products are not as clear as the distinction between species.
Living organisms share a common purpose, to perpetuate their being (i.e., genetic composition) into subsequent generations (in the case of kin selection, it may not be necessary to pass on an individual's specific genetic composition, but that of a close kin is sufficient). For web products, generally, the common purpose is to perpetuate their existence through sustained usage. This requires that web products attract first-time users and then, maintain continued usage.
The ability of living organisms to perpetuate their genetic make-up varies across individuals. This is true for individuals of a given species, as well as different species in competition with one another. Variability exists in the ability to access and utilize essential resources (e.g., water, sunlight, nutrients, territory, etc.). This is the basis for "fitness." Certain individuals are better equipped to perpetuate their genetic make-up and thus, are considered fitter than less able counterparts.
For various web products, the resources enabling perpetuation may differ. Generally, the attention of users, synonymous with time, is a sufficient resource. However, with e-commerce, since cash-flow may be necessary to continued existence, revenues are an essential resource.
A given ecological system may be represented by a 3-D topographical landscape. The y-axis indicates the degree of fitness, or the ability to perpetuate one's genetic make-up. The x and z-axis combine to represent hypothetical strategies for extracting and utilizing essential resources. For different ecological systems, the ruggedness of the landscape may differ, but in general, a series of hills and valleys may be anticipated. Those individuals and species at higher points on the landscape will be expected to perpetuate themselves. Individuals and species at lower points have a lower likelihood of perpetuating themselves, and where the individuals of a species are congregated within a valley, that species risks extinction.
Similar landscapes may be used to represent the Internet, or an intranet. The ability to attract and sustain usage may be represented on the y-axis. The x and z-axis would represent hypothetical strategies that could be employed by web products, perhaps different combinations of the factors that have been shown to predict usage in our earlier studies.
Shortcomings in Earlier Conceptualizations
In previous sections, discussion emphasized a static, Darwinian interpretation of ecological fitness. Assuming a static universe, one might anticipate that living organisms, or web products, would evolve over time till all were concentrated at the peaks of the landscape. In reality, fitness landscapes are not static. For example, where resource utilization exceeds resource renewal, resulting in a net loss in essential resources, strategies that rely on heavy consumption of available resources will be successful at first, but not after exhaustion of essential resources. Occasionally, the landscape may be shuffled and strategies that were once at a peak may drift to lower points, even valleys. Consequently, the fitness of living organisms, or web products, that have committed to those strategies will diminish perhaps resulting in their extinction.
Furthermore, in addition to gradual shifts in the landscape, natural systems are also susceptible to stochastic events that produce dramatic shifts in the fitness landscape. The meteorite collision that led to the demise of the dinosaurs and the majority of other species living on earth at that time illustrates the effects of a stochastic event on ecological systems. In these situations, there is a radical upheaval in the landscape and most living organisms, or web products, will not survive. In fact, those who had been the fittest may be replaced by those who had been the least fit.
In our earlier conceptualization of web usage, we had assumed a relatively static landscape. This assumption is evident in the notion that a collection of factors captured in a mathematical equation could be identified that would predict web usage over an extended period of time. Instead, the specific factors, their exact influence, and the interactions between factors is believed to be somewhat unstable, and evolving as the technology, and associated culture evolves. The best news site today, without change, may not be able to sustain usage one year in the future, if there are substantial shifts in the technology and culture associated with the Internet. Consequently, as has been noted elsewhere, it may be advisable to not over-specialize since over-commitment to a given strategy may lessen one's ability to abandon that strategy when the landscape shifts (Kelly, 1998).
Conclusion
The intent of this paper has been to place our earlier research in a slightly different perspective. We believe this perspective, an ecological model, offers greater insight into the dynamic properties of web usage. Much more could be said about the ecological model (e.g., hybridization, isolation of local peaks, etc.). But in closing, one point seems worth mentioning. A constant criticism of ecological models has been the difficulty of predicting events, as opposed to explaining events after their occurrence. This is largely a consequence of the slow progression of events with ecological systems. In contrast, most would agree that the Internet is evolving at an extremely rapid pace. Consequently, the Internet offers unique opportunities to develop and test ecological hypotheses within a reasonably brief time frame using a large-scale, dynamic system.
References
- Grose, E. & Forsythe, C. (1998). Predicting the penetration of intranet products: Developing the model. Internetworking 1.3.
- Kelly, K. (1998). New Rules for the New Economy. Viking, New York.
© Internet Technical Group
Last update: December 19, 1999
URL: http://www.sandia.gov/itg/newsletter/dec99/dear_aeby.html
hosted by Sandia National Labs
Disclaimer: Neither Sandia Corporation, the United States Government, nor any agency thereof, nor any of their employees makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately-owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by Sandia Corporation, the United States Government, or any agency thereof. The views and opinions expressed herein do not necessarily state or reflect those of Sandia Corporation, the United States Government or any agency thereof.