The media profession seem to agree there needs to be a rational way to put a price on Web Site Advertising. One Internet advertising agency, ModemMedia, has proposed a pricing model which has received considerable attention (Internet World , May 1995, p.59). There needs to be an examination of the ModemMedia pricing model from the Media Planner/Buyer's perspective. It is important to note that this model appears to be based on advertising which is designed only to bring people who are browsing somewhere else to the advertisers' own web sites by use of an ad which is a clickable billboard connected to a Web site which is the true ad. A cover blurb plus a full story, so to speak. Yet, the traditional media analogies used in support of the model range from niche magazines to billboards to direct mail.

ModemMedia's model has three elements:

• Determine the ratio of hits between the web site's log and the number of file "hits" that make up the page carrying the ad. Divide logged hits by number of hits making up the page to calculate what we can call "page views." Then call page views "reach."

• Determine repeat viewing of that page and call that frequency.

• Determine the success of viewings of that billboard ad in moving readers to the actual web site and call that "depth."

For use by media buyers, problem areas include terminology, calculations and value basis. The ModemMedia model apparently applies niche magazine CPMs of $70 to what are more billboard-like impressions. A CPM of $2 might be more typical for outdoor posters. And the model then applies a direct mail cost-per-person-opening-the-envelope to those who enter an advertised site. It also counts as impressions repeat exposure to the same page" in one reading session. Magazine CPMs, of course, don't count more than one impression per reader per monthly issue, no matter how often the book is picked up or opened.

TERMINOLOGY

The terms as well as numerical formulae need to be understood. Reach is a standard media measurement. It refers to the number of unique, different people exposed to a campaign or schedule of advertising. This may be expressed as a whole number or as a percentage of the total population being measured. In most media--broadcast or print--this "reach" of a single ad is equivalent to the rating of a single unit of advertising: a 30 second commercial or page in a single magazine issue. This equivalence is because there is a single defined time period associated with the ad unit. A viewer of a single commercial or a reader of a magazine issue is only counted once by the standard audience research used by media buyers (even if a reader picks up a magazine issue 3 or 4 different times). Visits to a Web site are somewhat analogous to magazine audience impressions, not reach (see below). That is, the computer process of counting hits--as used in the ModemMedia model--deals with the number of time a page is viewed. It does't distinguish between 5 people each visiting once (reach=5, impressions=5) and one person visiting 5 times (reach=1; impressions=5). Some have suggested that this should be a parallel to magazine circulation, but that analogy doesn't work: circulation is the number of copies available for reading--it ignores how many times each could be read and is a static figure. Web site availability is virtually infinite, limited only by the numbers of people who are equipped to read the content of a site. Impressions should be counted as one per visitor per month, or per whatever the time cycle of changing site content may be. This would then create a true parallel to other media.

The dynamics of web advertising are substantially different. It's as if a magazine had a device sensitive to touch and all copies of a magazine were collected after the issue life was over and the touch meters tabulated. We wouldn't know who had touched or how often, only the number of touches. It's also as if this magazine meter ticked over every time the eye move to a new article on the same page as an ad, and ticked again if the eye noted a picture accompanying the article. That, more or less, is how Web site hit count works.

FORMULAS

There are three essential formulas in the ModemMedia Model:

• A reach estimate (monthly hits divided by number of hits used to load a page), which as discussed above is actually an impressions estimate.

• A frequency estimate, which also turns out to be an impressions estimate.

• A depth estimate, which charges for readers as if they were sales.

The ModemMedia formula starts with a way to estimate hits, ie: watch the lower left corner of your Web Browser screen as a Home page is loaded by your computer. In that corner is a counter which records the bytes of the file being received ("xxx" of "yyy." As the "yyy" number changes, it indicates a different file being accessed).

A single page may consist of 5 or 10 files: The text, plus various graphics (logos, photos, section headers, up or down buttons, mail box buttons, "click here" boxes). So, since the term "hits" refers to a count of each file used in a page and the advertiser cares about the page as a whole, ModemMedia's idea is to watch the changes in the byte counter. Then add returns to the top page from other parts of the site as hits, and divide the total hits in loading the page or returning to the top page into the monthly hit count reported by the site. This then, says ModemMedia, gives an estimate of reach.

PROBLEMS

For a start, the browser alternates between the files in loading the page. On my browser (Netscape, the industry standard), the counter in the corner appears to show that the downloading process alternates between files, thus there are more hits apparently being displayed there than the number that a Web site log would report. Hence, ModemMedia's formula would have the potential to underestimate actual visits.

The formula counts returns to that top page as one hit to be added to the divisor by the formula. However, most Web browsers cache Web pages and documents, and so a return to the page in question would very likely be--in actuality--a page retrieval from the browser's cache, not an additional access to the site. This means that a return to the page during a single visit might well not add to the site's hit count at all.

This then causes further underestimates of actual visits.

FREQUENCY

The ModemMedia "frequency" estimate is based on returns to the home page (presumably this is the page where the billboard ad is). This involves multiplying the so called "reach" estimate by the previously estimated number of returns to the home page. Again, this would actually be an impressions figure. Note that "reach" above already counted these impressions, though now they are discounted by two-thirds before being applied to the $70 magazine CPM.

One justification cited for acceptance of ModemMedia's generous audience counts is reference to the industry's crediting every ad in a magazine with the entire reader audience measured, even though not every reader may see every page. By the same token, a web browser returning to the typical home page on which the billboard ad was located may not see that ad on the PC screen which displays less than the full page in one screen. As stated above, impressions should be counted only once per visitor. While I have pointed out some minor undercounting of visits in watching hits on the screen corner, the overcounting of impressions caused by including revisits to the same page on the same trip is a more extreme exaggeration of impressions.

DEPTH

Now, ModemMedia suggests that a direct mail value model of $1 per piece which is opened should be applied to those who click on the ad, which they estimate to be 20% of those reached, and go to the web site being advertised. This is based on an average price for a direct mail piece of 75 cents and 75% opening of the mailing.

PROBLEMS

Mechanically, clicking on a link to an advertised web site seems like a reasonable analog for opening a piece of direct mail, but there are other issues.

Leaving aside for the moment whether or not the 75 cents and 75% are realistic, there is an inherent illogic in the formula suggested, of 75 cents/75% = $1 value. If only 50% of the recipients opened the mail, then the calculation becomes 75/.50 or a value of $1.50. In other words, this would dictate that the less often a direct mail piece gets opened, the more valuable it is: a confusion of cost and value. ModemMedia's formula makes the weaker advertising more valuable. To take the logic to its extreme. In a mailing of 75,000 pieces, if only one was opened, that one has a value of $5.6 billion dollars. Yet if one less than that is opened, clearly the total mailing has a value of $0. By the way, direct mail advertisers typically track response, which may be 5%, rather than "openings."

All this is meant to suggest is that using this arbitrary direct mail valuation has no relevance to the web site ad. The value of direct mail is sales generated per piece, not readers.

Finally, it is suggested in the model, that an estimate of 20% of those reached by the billboard ad will click on the link and enter the advertisers website. Therefore, the above $1 cost per opening is applied to the estimated 20% of 10,000 reached who will "click", for a total of $2000. This $2000 is added to $700 for the 10,000 reached by the clickable billboard at a $70 magazine CPM, plus one-third of the repeat links to the home page (.333 x 10,000 x 4) for $930 for a total price of $3630, for an ad with an estimated 2000 readers, or a CPM of $1815! Or, to be most fair, $1630 being charged for the 10,000 "reach," creates a $163 CPM. If you accept the discounted 40,000 return impressions, which of course shouldn't be counted in a CPM, you have a $70 CPM for impressions which are most comparable to outdoor billboard impressions. Which leaves the $2000 for the 2,000 actual readers of the Web site for a $1000 CPM.

SUMMARY

Bottom line, the Modem Media model seems to over-count impressions, call them reach, and justify CPMs of $1815 for an audience whose quality is only assumed. Web ads are magazine-like, not direct mail-like, unless they generate orders, which may well make them worth $1815 CPM, but at least will make it easy to put a value on them.

SUGGESTIONS?

The problem is in trying to incorporate production cost into media when the proportions don't work. Internet advertising agencies need to sell their creative costs separately. From a media perspective, the ModemMedia model has moved to the point of rational thinking. All it needs is to count impressions in CPM calculations in the same way as other media do and to put values on those impressions comparable to the media they are most like. For "billboard" impressions, something between outdoor and magazine CPMs will probably be acceptable to the industry, and once per visit to a site is the fair count of impressions in those CPMs. If clicks are sales as in true direct marketing, then another model, perhaps commision based, may become appropriate.

Abbott Wool is a veteran of 25 years in advertising. At industry-leading firms like Saatchi & Saatchi, Vitt Media International, Foote Cone & Belding, and Strategy Research he has been a Media Director, Media Research Director, and Ma rket Research Director. He operates the Market Segmentation Resource Locator.