1 2 3 4 5 6 7

genuine access only to a small subset of employees. There are also subtle but real distinctions to be made between types of Internet access: indirect access via a gateway from an online service or bulletin board system (BBS) gives you a more restricted set of functions than direct access via an Internet service provider (ISP) that puts your machine directly on the Net; direct access may or may not give you your own domain name and permanent numbered Internet address, allowing you to host a Web or FTP site on your own machine instead of on a service provider's setup.

The O'Reilly survey, with the lowest numbers, defined an Internet user as someone eighteen or over with direct Internet access, excluding users whose access was solely via an online service such as A O L or CompuServe. FIND/SVP included all those eighteen or over who used at least one Internet application in addition to email, with no restrictions on whether that Internet access was supplied by an online service or a direct ISP. Times Mirror's substantially larger estimate was based on the broadest definition: anyone eighteen or over who ever used a computer from home, work, or school to connect to online services, BBSs, or the Internet.

Raw Internet usage was not the only element these surveys studied; they also looked at the overall demographic makeup of those who use the Internet, including g e n d e r, education, income, and age. Stanford Research International (SRI) analyzed the Nielsen and O'Reilly surveys along with two sets of figures from the Graphic, Visualization, and Usability (GVU) Center at Georgia Tech looking for overall trends. Among other things, SRI found that the percentage of women on the Net was rising; that the average age of Internet users was likely to decline, slowly dropping to slightly below a median age of thirty; and that average income of Internet users might also decline. SRI concluded, however, that "educational attainment continues to be the fundamental demographic driver of Web access and a c t i v i t y. "[10] Its prediction that women would make up half of all W e b users by the end of 1996, however, has not come true; best estimates at the end of 1996 were that a little over a third of Internet users were women. Of the areas SRI studied, it found the most difficulty in drawing conclusions about income, which, they note, "remains an ambiguous and unreliable measure of the value of the Internet audience."

Darrell Huff, whose short, entertaining book How to Lie with Statistics ought to be required reading for anyone who is bemused by the many numbers we see in newspapers every day, wrote, "Many a statistic ... gets by only because the magic of numbers brings about a suspension of common sense."[11] Using a common-sense yardstick, most Internet users probably thought Nielsen's estimates were high, just as most experienced Usenet posters know instinctively that Rimm's claim that 87 percent of Usenet traffic is pornography can't be right. Personally, the most valuable of these surveys is the SRI meta-analysis, which looks for trends rather than focusing on specific numbers.

In his analysis of the Rimm pornography study (see chapter 9) DEC research scientist Brian Reid goes on to point out that useful information can be derived by studying trends--for instance, comparing the results over time for a single newsgroup, or comparing the results from a single time period across newsgroups. While Reid's comments were limited to Usenet, a very different ball of wax, until we have independent analyses of Internet usership that are non-commercial and equivalent to the decennial federal population census, bearing these principles in mind seems to me a good idea. Although, pretty much all the surveys agreed, with the exception of Nielsen. Time will also help; a few years from now, companies that have been selling via the Web, such as L. L. Bean, Land's End, and the bookselling service Amazon, will begin to report the results of their experiences.

It is, of course, in the interest of anyone selling the Internet as a commercial medium to show that there are lots of women (the primary shoppers in many households), high income levels, and all those other things that advertisers like. SRI's caution, in the midst of so much Internet hype, is therefore particularly welcome.

In the end, what constitutes success for advertisers is people buying their products. Measuring this kind of success on the Web turns out to be no easier than