It's a valid point. If you're going to market effectively, you have to understand your marketplace. That principle, (Harold Hill's experiences in River City notwithstanding,) is just as valid for virtual marketing as it is for marketing in the "real" world. As the businesses for which we design and manage networks increasingly move to adopt the World Wide Web as a marketing venue, the success or failure of that shift depends in large measure on how well we understand the makeup of the "virtual territory" of the Internet.

Luckily, there are now several relatively scientific studies of Internet demographics, marketing trends and user preferences available on the Web. Unfortunately, these studies don't agree very well about how to define an Internet user and they also differ substantially in methodology, so whose results you choose to believe will depend on which definition and methodology you most favor.

The Graphics, Visualization & Usability Center (http://www-survey.cc.gatech.edu/cgi-bin/Entry/) of the College of Computing (http://www.cc.gatech.edu) at Georgia Institute of Technology (http://www.gatech.edu) in Atlanta, GA, began its Fourth WWW User Survey (http://www.cc.gatech.edu/gvu/user_surveys/) on October 10, 1995. The survey was conceived, developed & deployed by James Pitkow, (http://www.cc.gatech.edu/gvu/people/Phd/James.E.Pitkow.html or mailto:pitkow@cc.gatech.edu) a GVU Graduate Researcher, and is being conducted by he, Colleen Kehoe (http://www.cc.gatech.edu/grads/k/Colleen.Kehoe/home.html or mailto:colleen@cc.gatech.edu) and Laurie Hodges (http://www.cc.gatech.edu/gvu/people/Faculty/Laurie.B.Hodges.html or mailto:laurie.hodges@gtri.gatech.edu). It covers the average age, gender ratio, average and median income, location, occupation, and willingness to pay for access to Web sites of an international sample of users. It also studies how often users browse the Web, what they use it for, (entertainment, information, direct purchase, etc.) and what services (ftp, Gopher, Usenet, native HTML, etc.) they use and what percentage of their total use each service comprises.

GVU also surveys information providers and HTML authors. GVU asks how easy it was for authors to learn HTML, how they learned it and which Web server software they use.

The GVU surveys are conducted via links to survey forms on very popular www pages (such as NCSA's What's New, Hotwired, Lycos, etc.) and posting survey forms to Usenet newsgroups, (comp.infosystems.www.*, comp.internet.net-happenings, etc.) at equal intervals during the month the survey runs. They also made an announcement at the Third International WWW Conference in Darmstadt Germany and they sought write-ups in computer- and Internet-related trade magazines and in daily newspapers. For the last survey, they also placed a link on Prodigy's Web access page.

The weaknesses of this survey method are that all respondents self-select, all reports are anecdotal, no attempt is made to confirm the accuracy of user-supplied information and all users are required to be English-literate in order to respond. Those who answer the survey are exclusively those who have heard of it, think it's a reasonable use of their own time and who have the time available to respond. The strength of this method is its international scope and the very large number of responses (GVU's Third Survey, which concluded in May, 1995, received over 13,000 responses.)

GVU shares its raw data with the Hermes project (http://www.umich.edu/~sgupta/hermes/), run by Sunil Gupta (mailto:sgupta@umich.edu) at the Michigan Business School at the University of Michigan. The Hermes project concerns itself with commercial usage of and preferences and attitudes about the Web. Its analysis of the results of the GVU Third Survey (including pie charts and bar graphs) is available in both Adobe Acrobat (http://www.umich.edu/~sgupta/hermes/survey3/survey3l.pdf) and PostScript (http://www.umich.edu/~sgupta/hermes/survey3/survey3a.ps) formats.

On September 28, O'Reilly and Associates of Sebastopol, California released the results of the survey they call "Defining the Internet Opportunity", (http://www.ora.com/survey/) which they conducted in cooperation with Trish Information Services (http://www.ora.com/survey/trish.html) of Hayward, California. This telephone-based survey purports to define the "true number of US Internet users, broken down by level of use, including email and tools such as FTP, gopher, Telnet, and the World Wide Web."

The O'Reilly survey was conducted in three phases. Phase 1 consisted of a mail-in survey of 2,052 users of the Global Network Navigator service, conducted in the days before O'Reilly sold GNN to America OnLine. Phase 2 involved randomly dialing U.S. telephone numbers (so as to include otherwise unlisted numbers) until a total sample of 1,000 current Internet users was created. In Phase 3, the same method was used to contact 500 users of commercial online services. There were, in O'Reilly's estimation, about 1,000,000 users who fit into both categories. In both Phases 2 and 3, only users who completed a 35-minute questionnaire were counted in the sample. In order to qualify the 1,500 respondents, Trish dialed around 200,000 telephone numbers and screened 32,000 potential participants.

There are a couple of weaknesses in the O'Reilly approach and at least one real strength. The strength is that, much more so than the GVU studies, Phases 2 and 3 of the O'Reilly surveys managed a reasonable approximation of truly random sampling. The weaknesses include O'Reilly's definition of an Internet user and the selection of only those who agreed to spend 35 minutes answering the questionnaire as part of the study sample. For purposes of the study, O'Reilly defined an Internet user as someone whose access to the Internet was strictly via an Internet service provider (i.e.--not via a commercial online service.) The problem with this approach is that all the commercial online services now offer more-or-less transparent access to the Web, (although, admittedly, most subscribers cannot yet make secure purchases via the online service's browser,) so I question its validity.