Through its explosive growth in recent years, the Internet has long been regarded as the antithesis of conventional broadcasting media such as newspaper, radio, and television. Instead of pushing generic content designed for mass consumption to the user, it lets the latter control the process of communication by pulling in only items of personal interest. The liberating effects of this approach appeared to culminate with the proliferation of the World Wide Web as the increasingly popular medium to publish and browse information. The potential for business applications in electronic commerce [11] was quite obvious. Indeed, as enterprises large and small scurried to stake their presence on the Web, hopes were high that once that is done, the whole world will come knocking. To date, this has rarely been the case [6]. In the haste of jumping on the band wagon, many commercial Web projects suffered from murky definition of business purposes and dubious value-added for the customers. See [10] for a global study of commercial web sites using a purpose-value framework from the perspective of consumers. In any case, with millions of sites vying for attention, advertisers were losing patience with the "pull" format. In came "push" technology to the rescue by delivering more or less customized content to a browser's desktop along with unsolicited advertising ([3], [12].) To the critical observer ([4], [13]), this approach is tantamount to retrofitting the Internet with cable television. While convergence of the two media may indeed facilitate penetration of the mainstream market, real progress is unlikely with the simple blending of one into the other. It now appears that such progress may not be totally reliant on the invention of more techno-gizmos. Rather, it can depend more on the creation of value [9] through truly compelling uses which take full advantage of the defining features of the Internet with a suitable balance of push and pull.

This work addresses one potential example of a compelling use known as focused broadcasting. It is essentially a matching service for classified and want ads. After explaining the mechanics of this model and where it fits in the push-pull framework of information dynamics, a pilot study in Internet user attitudes toward such service is presented. The study, conducted in June 1997, involved five groups of Internet users in three cities in New Zealand, a country that mirrors the rapid development and deployment of the Internet infrastructure worldwide. Conjoint analysis was used to identify user attitudes toward three attributes in the model as well as cluster groups of various user types. It was found that across all cluster groups, there is consensus on the appeal of the approach. Attribute partworths and importance weights obtained from the analysis can be used to evaluate specific design options in actual planning and implementation of focused broadcasting on the World Wide Web.

Information Dynamics

To capsulize the ongoing debate over the relative merits and potentials of push versus pull technologies for the Internet, it is helpful to format. In came "push position various conventional and emerging media as shown in Figure 1.

Figure 1. Schematic of Information Dynamics

A two-dimensional framework is used. Horizontally, the relative characterization of how information is disseminated and acquired is displayed. Note that to the extent that information is being made available for public consumption, there is always an element of push in the sense of broadcasting, whether it is in print, air waves, or electronic signals. Without any attempt at formal definitions, it is important to have some guiding principle to distinguish push from pull. We propose a simple one in the form of a question: "Can you look it up at your own leisure?" If so, then it is pull. This is certainly the case with books in a library or newspapers on a coffee table, but not so with the real-time broadcast of radio or television programs. The vertical axis indicates whether the content of the medium is primarily designed for mass consumption, or customizable to individual interests. The question to ask here is "To what extent do you see only what you are looking for?" In this sense, the distinguishing emphasis is put on how an individual is guided to information of interest, rather than the selectiveness of the overall material. For example, a typical book is a line by line presentation (linear text) designed by the author. Much as any reader can skip around the pages, the layout is the same for everyone and primarily meant to be followed cover to cover. By contrast, a dictionary, while providing the same overall content to every user, is specifically designed to help find only what one is looking for.

Within this framework, both conventional and emerging media can be characterized. Book and newspaper are highly mass-pull while radio and television are highly mass-push. Bulletin boards, especially those with classified and want ads, sit higher on the custom scale. (The same goes for special sections of newspapers for this purpose.) So-called 1-to-1 marketing [17], in which individual purchasing behavior is culled from massive customer databases and used to target advertising campaigns can be considered highly custom-push. Three Internet-based media are shown in italics. Online services, in the role as content providers, are somewhat more custom-pull then newspapers. The World Wide Web, as a repository of hyperlinked multimedia, is much higher in the same column. Webcasting [3], which has gained much attention recently, is push and not much higher than online services on the custom scale.

The purpose of our framework for information dynamics is to provide a perspective for future development of Internet applications. Observe that mass-pull and mass-push are already saturated by conventional media, and those emerging to-date tend to fall into the same push-pull divide [20]. This suggests various unexplored territories and the prospect of truly compelling applications bearing the corresponding characteristics. In the following, we introduce one such example which is a highly custom balance of pull and push.

Focused Broadcasting

Much as the Internet is undergoing explosive growth, penetration of the mass market is still far from reality [5]. It is quite conceivable that significant breakthrough may not depend solely on further advances in the technology, but rather on truly compelling applications of what is already available. Take senior citizens for example. Most of them have no interest in high technology in its own sake. Judging from demographic data (see e.g. [8],) efforts to pitch computers to them as gadgets of pastime or tools to keep an active mind have only very limited success to date. However, grandparents are likely to find "videoconference" over the Internet with their out-of-town grandchildren compelling [15]. Others may be similarly impressed by healthcare resources on the Web. In our context, any feature that is attractive enough to lure people into gaining access to the Internet is considered truly compelling.

We now introduce a model that has a promising niche in our framework of information dynamics. The term Focasting is trademarked by Infotomics Inc., a Wilmette, IL-based consultancy, and used with permission throughout this work. It stands for "focused broadcasting," an advertising system that takes full advantage of both the culture and the technology of the Internet in general, and the World Wide Web in particular. Three parties are involved: Viewer, Focaster, Advertiser. The Viewer registers with the Focaster (who may be an ISP, E-Mall, Directory, Web-Host, etc.) free of charge, indicating specific items, categories, or sources of interest. Provided with an ID and Password for subsequent visits, the Viewer sees Web pages at the FocasterŐs site showing only timely information furnished by Advertisers of interest. The Advertiser pays the Focaster at rates that can be meaningfully established and negotiated based on Viewer demands.

Focasting is essentially an automated matching service for classified and want ads over the Internet. However, it is not information brokering, purchase agency, or any traditional role of go-betweens and middlemen. As the trend in cyber-culture is toward disintermediation, even automation of such models is not promising. Instead, the Focaster strives to be an efficient conduit of valuable information. Apart from helping to classify and match the supply and demand, no value judgment is exercised. Hence, no value-based commission is charged. The Viewer sees only what is of interest, for free. The Advertiser reaches only those who are interested (a focused audience), and more cost-effectively than with conventional advertising. The Focaster earns a fee by providing the service, leveraging Web technology for economies of scale. Why not charge the Viewer a fee? Obviously, the Viewer derives value from the service and may indeed be willing to pay for it. However, competition among Focasters will eventually drive this fee to zero. As long as Focasting generates sales for the Advertisers, they have an incentive to pay.

There are many precursors to the building blocks of the Focasting system. Customized reports and newspapers embody the same technology, but not the economics. Existing E-Malls and Industry Networks serve the same purpose, but lack the dynamics to balance the pushing and pulling of information. Ultimately, it is the recognition that by catering fully to the Viewer, the Focaster ends up with a product that is perfectly marketable to the Advertiser. With the new focused channels, the Advertiser can concentrate on timely and customized announcements, rather than the relatively static and mass appeal approach in conventional media.

Figure 2. Information Dynamics of Focasting

The relative positioning of Focasting in the schema of information dynamics is illustrated in Figure 2. Note that it can be regarded as a balanced act of pull and push since contents of interest are linked in real time and pushed to the Viewer. However, the latter takes the initiative and control of the timing of communication. The only basic technical requirement to implement Focasting is the ability to generate data-based Web pages on the fly, which is already widely available with most server systems. There is no drawback in terms of high bandwidth demands as with existing push technology.

While obvious applications abound in real estate, collectibles, automobile, travel and leisure, as of mid-1997 only a few commercial Web sites have features that fit the Focasting model. The prime example is the Personal Search Agent service offered by The Monster Board (www.monster.com,) a job search/recruiting site. Here the registered Viewer may maintain up to three profiles, each indicating a choice of job type and geographical location. On subsequent visits, the latest job postings by Advertisers that suit the Viewer's profiles are displayed. Another example is Cyber-Merchant Exchange (www.c-me.com,) an electronic trade-show site for wholesalers and volume buyers, that has formally adopted the Focasting model [1].

A Pilot Study in User Attitudes

To explore the potential of Focasting as a compelling application that may play a role in fostering electronic commerce, a study of Internet user attitudes toward the model was conducted. The opportunity arose during the authors's visit to New Zealand in May and June 1997 as a Visiting Erskine Fellow at the University of Canterbury at Christchurch. We begin by explaining why this backdrop is particularly suitable for a pilot study that can produce results indicative of user attitudes worldwide.

Environment

With a population of 3.6 million, New Zealand is among the top countries in the development and deployment of the Internet infrastructure. According to statistics compiled by the Ministry of Commerce [18] in 1997, one out of four households has a home computer. The number of Internet host computers increased from 1,193 in July 1991 to 84,532 in January 1997, a growth of 700% compared to that of 300% worldwide. In terms of the number of Internet hosts per 1000 people, New Zealand ranks seventh in the world with 23. The top ten countries are led by Finland with 56. The US is fourth with 38; Canada is ninth with 20. As of March 1997, there were 9,620 registered Internet domains in New Zealand, 8,343 of which being commercial. In 1996, the mean student to computer ratio in the 2,736 primarily and secondary schools was 18. About a third of the schools were connected to the Internet, with most of the rest intending to do so by January 1998. A directory of ISP's had 64 listed as of June 1997.

Given these parameters, and the fact that it would be difficult in any case to arrive at truly representative results of a global nature, we believe the results will be useful as a significant pilot study.

Attributes of Focasting

Gordon and De Limer-Turner [7] studied consumer attitudes toward Internet advertising in general. They pointed out that unlike other forms of advertising, the social contracts governing Web advertising is evolving and not well understood by all parties affected (consumers, advertisers, media, and agencies.) Primarily concerned with how unsolicited advertising is pushed to the consumers, they identified several attributes as the subjects of study: access, placement, message, influence on fee, and information collection. For Focasting, the business model is better defined as a balance of soliciting information (pull) and advertising (push) so that most of the above attributes are no longer relevant. Still, using the same perspective of social contract, the following attributes of Focasting are found to be important in its implementation. While further refinements are certainly possible, we consider only two levels for each of the three attributes in this study.

i) Communication

This attribute reflects the push-pull balance. Level zero is our base model, without any intervention on the part of the Focaster. The Viewer logs on the Focaster's site to view latest information at will and without prompting. Level one involves notification by e-mail whenever there is new information of interest. While this feature, which elevates the degree of push in the model, may appear to enhance the timeliness of communication, in practice it can multiply the volume of e-mail and eventually become a nuisance.

ii) Latitude

This attribute controls the strictness in adherence to customization. At level zero, the Viewer sees only ads that match exact criteria stored in the requested profiles. For example, if one is looking for a used car of a specific make, age, condition, and price range, then only ads for such cars will be shown. At level one, the Focaster accepts additional sponsorship to increase revenue. For instance, an ad for an auto paint shop may be included for any Viewer looking for a used car. Unlike totally unsolicited ads, such related ones may be acceptable to many users. With such sponsorship, the Focaster can offer more affordable rates to the focused Advertisers and hence improve on the service to the Viewers.

iii) Privacy

In the specification of an interest profile (the want ad), the Viewer can be offered a selection of options that are totally impersonal. For example, in a job search profile, only the type of job, location, etc. may be used. The matching is then performed to gather what the Viewer is looking for. This is level zero where no other personal data need be divulged. At level one, additional demographic data (gender, age, qualification, experience, etc.) are involved. This may help generate better matches of potential interest.

Sample

The selection of participants is a cross between convenience sampling and focus groups. The entire sample comprised five separate groups. The first included 35 students in a third-year course in telecommunication systems. The second had 4 students in a graduate research seminar. The third group without prompting. Level one in was from a large class in a first-year introductory course in information systems and technology, in which 110 students participated in the survey. These three groups were from the University of Canterbury at Christchurch. The fourth group had 27 graduate students, mostly practicing professionals, in a diploma program in information management at the Victoria University of Wellington. The fifth group involved 14 participants, also mostly practicing professionals, in an executive seminar at Auckland University. The total number of participants was 190.

Being affiliated with universities and students in information technology related courses, all the participants are users of the Internet. The author's access to the groups as guest lecturer made them convenient samples. An average of approximately 40 minutes was spent in surveying each group. A brief introduction of the push-pull dynamics, the Focasting model, and a demo of the Monster Board case were presented and discussed as in focus groups. Finally, a written survey was completed asking for the likelihood of the participant using Focasting under various scenarios.

All five surveys were conducted between June 16 and June 27, 1997. Although our sampling was unconventional, the resulting demographics suggest that it can very well be representative of the regional population of Internet users at large.

Instrument

With three attributes at two possible levels each, there is a total of eight full profile scenarios. Since this should be a manageable number of tasks for the participants, all were asked to rate the likelihood of their using each scenarios. A 7-point scale was used, with (7) indicating "definitely," (6) "very likely," (5) "somewhat likely," (4) "undecided," (3) "somewhat unlikely," (2) "very unlikely," and (1) "definitely not."

Demographic, PC ownership, Web usage, and questions on experience with buying and selling through classified and want ads were also asked. Ten to fifteen minutes were allocated to completing the survey.

Analysis

The approach of conjoint analysis [16] was used to identify user attitudes toward the three attributes in the model as well as cluster groups of various user types [14]. First, each participant's responses to the full profile scenarios were fitted in a linear regression using dummy variables. Levels zero and one of each attribute were represented by an independent variable at value of 0 and 1, respectively. The dependent variable was the likelihood rating. The coefficients of the regression were interpreted as estimates of the partworths for the corresponding attribute level.

Since the setup for the study is far less elaborate than that for the typical focus group, it is necessary to screen for usability of the responses. Given the limited time frame, some participants may not have understood one or more of the elements: the Focasting model, the scenarios, or the survey task itself. As a convenience sample, some participants may not have genuine interest in taking the task seriously or forming a thoughtful opinion. Assuming that such cases were likely to lead to inconsistent responses, we used the R-sq as a measure and rejected cases with a value under 0.5 for the analysis. Of the 190 responses, 147 were retained with a mean R-sq of 0.8.

Next, to identify cluster groups of user attitudes, standardized partworths obtained from the regression coefficients for individuals were analyzed using Ward's method as implemented in SPSS for Windows (Release 6.1). Preliminary examination of the results for 2, 3, 4, and 5 clusters suggested that the 3-cluster partition had the most meaningful interpretation. As the hierarchical approach in Ward's method did not allow reclassification of cases, the 3-cluster partition was refined with discriminant analysis. Stable clusters were obtained after reclassifying seven cases in the first iteration and two cases in the second.

To compute attribute importance weights for each individual case, the difference between the regression coefficients of the two levels for each attribute was divided by the sum of such differences for all three attributes. Note that there were a number of cases with zero regression coefficients for all attribute levels. These had to be interpreted as having "equal" importance weights for each of the three attributes for the consistency of subsequent analysis. Demographic characteristics of the three cluster groups were compared using anova for continuous variables and chi-square tests for categorical variables. Mean partworths and importance weights were tabulated for the cluster groups as well as the entire sample. Anova was used to test for significance.

The highest rating among all eight scenarios by each individual is considered the likelihood of that individual using a Focasting service. The distribution of these likelihoods were compiled as an indicator of how compelling Focasting is as an application of the Internet. Chi-square tests were used for variation among groups.

Results

From the original 190 cases, the sample was reduced to 147 (henceforth referred to as the whole sample) after eliminating those judged to be unreliably inconsistent. Clustering and refining with discriminant analysis, three groups that were later labeled Realists, Purists, and Optimists were obtained, with membership of 36, 66, and 45, respectively. Their characteristics were summarized in Figure 3.

To identify variation across the cluster groups, Chi-square tests were performed on the group characteristics at the 0.05 level of significance. It is found that age distribution is marginally independent of grouping (p=0.057). Gender mix is independent of grouping (p=0.16). Web usage varies across groups (p=0.01). Ownership of PC's (p=0.73), buying (p=0.30) and selling (p=.19) experience with classified/wants ads are all independent of grouping.

The mean partworths of the attribute levels and the mean importance weights for the attributes by cluster groups are tabulated in Figures 4 and 5, respectively.

	SAMPLE	REALISTS	PURISTS	OPTIMISTS
Attribute	(N=147)	(N=36)	(N=66)	(N=45)
Communication	0.38	0.70	0.26	0.28
Latitude	0.36	0.13	0.55	0.28
Privacy	0.26	0.17	0.19	0.44

Figure 5. Mean Importance Weights

We now explain how the mean partworths of the cluster groups led to their nomenclature. The first group of 36 are, on the average, strongly negative (-0.7) about e-mail notification, marginally negative (-0.06) about related but unsolicited ads, and marginally positive (0.05) about supplying demographic data. We refer to this group as Realists as they appreciate how easily one can get overloaded with e-mail, that advertising sponsorship is perhaps an unavoidable economic reality, and that effective customization requires personal data. As Web usage is the only characteristic that varies significantly across the cluster groups, we observe that the Realists are the more frequent and hence more experienced users of the Web. The second group of 66 are quite negative (-0.18) about e-mail notification, very negative (-0.55) about related but unsolicited ads, and marginally negative about supplying demographic data. They are called Purists as they seem to be resisting the commercialization of the Internet. This is consistent with the fact that they have the highest percentage of occasional users. The third group of 45 are quite positive about all three factors (0.15, 0.13, 0.16, respectively.) We call them Optimists as they appear to embrace value-added without being concerned with the tradeoffs. If we do consider a slight variation in the age distributions among groups, the Optimists tend to be younger, with 73% under 26 compared to 57% for the whole sample.

Anova returned significant differences for all entries among groups except the following. Mean partworths for Privacy were not significantly different between Realists and Optimists. Mean importance weights for Privacy were not significantly different between Realists and Purists. Mean importance weights for Communication were not significantly different between Purists and Optimists.

For an indicator of how compelling Focasting is as an application of the Internet, the highest rating among all eight scenarios by each individual was used. This is the likelihood of that individual using some version of Focasting. The distributions of these likelihoods for the cluster groups are tabulated in Figure 6.

For the whole sample, 93% are at least somewhat likely to use Focasting; with 76% very likely or definitely. From a Chi-square test, the likelihood distributions are dependent on the groups (p=0.03). This is due to the relatively high number of "undecided" among the Optimists. Setting aside this category of response, the likelihoods are not significantly different among the cluster groups (p=0.9).

Discussion

The pilot study of Internet user attitudes produced strong evidence that Focasting, a model for matching up classified and want ads on the World Wide Web, can be a compelling application. Nearly all (93%) in our sample indicated likelihood to use it in some form. As to variations in user attitudes, three cluster groups were identified and interpreted as Realists, Purists, and Optimists. One important finding is that all three groups are of significant proportions, comprising 24%, 45%, and 31% of the whole sample, respectively.

Since both Realists and Purists do not care for e-mail notification, this feature, even as an option, can be assigned relatively low priority. The Purists, being dominant in number and less frequent users, represent the major challenge in penetration of the mass market. Their distaste for unsolicited advertising will have an impact on the economics of Focasting. Indeed, one way to cater to this group is to provide an option in their profile to switch sponsoring ads off. However, the reduction of exposure implies lower revenues from sponsors [19].

Another somewhat surprising finding is that across all groups, users are not as negative about providing input data to enhance matching as commonly perceived [2]. This is perhaps due to the understanding that useful data can be elicited for Focasting without becoming intrusive. For example, in a job search profile, area of expertise, previous positions, salary history, etc. can all be useful without actual names, addresses, or any data for positive identification. Again, the best strategy here seems to be flexible and offer as many options to the user as possible.

As limitations of our pilot study, we emphasize both the regional and convenient nature of our sample. Based on consensus analysis of myriad surveys and demographic studies (see e.g. www.e-land.com), a globally representative sample would probably have a higher mix of older users as well as women.

Finally, with the encouraging results from this pilot study, future research on more specific implementation of Focasting, such as for real estate, collectibles, automobile, travel and leisure, can be expanded along the lines of additional attributes and levels. This work provides the foundation for exploring a model that has the potential of a compelling application to foster electronic commerce over the Internet.

References

[1] R. Balu, "Buying and selling in cyberspace: a biz model," Crain's Chicago Business, 5/12/97.
[2] J. Bresnahan, "Up close and personal," CIO, 5/15/97.
[3] Business Week Special Report, "Webcasting,"Business Week, 2/24/97.
[4] D. Caruso, "When push comes to shove," The New York Times, 3/24/97.
[5] J. Coates, "E-commerce eludes on-line retailers," Chicago Tribune, 4/17/97.
[6] K. Essick, "Internet users slow to adopt electronic commerce," Computerworld, 4/14/97.
[7] M.E. Gordon and K. De Lima-Turner, "Consumer attitudes toward Internet advertising: a social contract perspective," International Marketing Review (forthcoming.)
[8] GVU's WWW User Surveys, Graphics, Visualization and Usability Laboratory, Georgia Institute of Technology, http://www.cc.gatech.edu/gvu/user_surveys/
[9] J.K. Ho, Prosperity in the Information Age: Creating value with technology -- from mailrooms to boardrooms, Infotomics (1994), http://www.uic.edu/~jimho/pia.html
[10] _____, "Evaluating the World Wide Web: a global study of commercial sites," Journal of Computer Mediated Communication, Vol. 3, No. 1, June 1997, http://www.ascusc.org/jcmc/vol3/issue1/ho.html
[11] R. Kalakota and A, Whinston, Frontier of Electronic Commerce, Addision-Wesley (1996).
[12] K. Kelly and G. Wolf, "Push! The radical future of media beyond the Web," Wired, March 1997.
[13] S. Levy, "The Internet gets pushy," Newsweek, 5/12/97.
[14] W. McDonald, "Internet customer segments: an international perspective," in C. Droge and R. Calantone (eds.), Enhancing Knowledge Development in Marketing, American Marketing Association (1996) 338-344.
[15] M. Meyerhoff, "The methods of a modern-day grandpa," Hemisphere, March 1996.
[16] J.M. Myers and E. Tauber, Market Structure Analysis, American Marketing Association, 1977.
[17] D. Peppers and M. Rogers, "One-to-one media in the interactive future," in E. Forrest and R. Mizerski (eds.) Interactive Marketing: The Future Present, NTC Business Books (1996).
[18] M. Templeton, "Statistics on Information Technology in New Zealand," Ministry of Commerce, New Zealand, March 1997, http://www.moc.govt.nz/comms/it/it97/
[19] D.A. Williamson, "Smart agents build brains into Net ads," Advertising Age, 4/8/96.
[20] G.M. Zinkhan and R.T. Watson, "Advertising trends: innovation and the process of creative destruction," Journal of Business Research 37 (1996) 163-171.