NONE: ONLINE-ADS>> Re: Real numbers on web ad / sales revenue
ONLINE-ADS>> Re: Real numbers on web ad / sales revenue
Harry Wolhandler (hwolhandler_at_activmedia.com)
Tue, 17 Sep 1996 22:03:24 -0400
Returning to the Ad Numbers Debate - -
Last week Donna Hoffman wrote re: Tracking Ad Dollars:
>If ActivMedia is not conducting a stratified random sample
>(note that they don't have any of the top ten sites in their sample!),
>then their results are not generalizable.
>It also makes one question the representativeness of your random sample, if
>it didn't include any of the largest revenue generating sites (assuming
>that WebTrack's top 10 are in some way related to the "real" top 10, of
>course.)
>I think we need to be very careful about understanding what can be "taken
>away" from studies that are not representative of the population they seek
>to describe.. =20
Perhaps there is a need to review the concept of stratified sampling:
establishing quotas to ensure proportionally correct representation of a
KNOWN population distribution within a representative sample.
The "Real Numbers" study was carefully stratified . . . by nearly 100
business subsegments with known population distribution who were doing
commerce on the web. Incidence of returns by business type were comparable
to incidence among the mailout sample. To sample a population by
establishing quotas for an UNKNOWN variable, as in the population income
distribution for all sites on the web, would render the results
unprojectable since the data are skewed. Just because a consumer study
doesn't include Bill Gates doesn't mean that it is not reflective of the
general population of the US. In fact, the study will be a more accurate
reflection of the population without Bill Gates. By adding Bill's income in
*after* we have projected from the rest of the sample, we gain a more
accurate estimate of total individual earnings than if Bill were included
in the sample, skewing the mean.
>Jupiter's Ad Spend may have bias, but it gives a reasonably good picture of
the "big picture."
Not at all! It is only a picture among the small minority of commercial
websites with posted advertising rate cards, and further only among those
listed in the WebTrack database. Clearly, if these 600 sites are your
audience, Ad Spend is valuable. And if you want to talk about "share of
adspace", using rate cards and display space as a proxy for income, you
would be delighted. But beware of making universal judgments about the
"big picture" of commerce on the web from this data! By design it
overlooks not only upwards of 10,000 websites with some form of paid
advertising or links (some with rate cards, per anecdotal evidence in this
newsgroup / some without - as in swapped or privately placed ads) as well
as many alternative forms of Net advertising (as in newsgroup visibility
management and organized marketing efforts to be widely linked).
Additionally, it relies on assumptions that are likely to be factual (rate
cards don't usually reflect actual negotiated final rates, and swapped
advertising is accorded the same value as paid advertising, for instance,
rather than being useful filler, as they are in many cases). This does
not invalidate their method or study design - - their study is simply what
it is - - an innovative form of adspace / value measurement.
>I think it's true that a lot is happening at the grass roots level, but
from
>an advertising perspective, major advertisers do not care at this point
>about buying space on Joe's Tire Site.
>Let's be clear about the objectives of collecting numbers before we start
>drawing conclusions about general market trends.
Following on the thread above, one must be careful to define just what the
study population reflects. In contrast to the focus of Ad Spend,
ActivMedia's "Real Numbers Behind 'Net Profits" is also what it is - - a
concerted effort to track growth of Commerce OVER the Web (as opposed to
measuring the commerce of the Web industry itself.) The intent is to look
at emerging businesses which are establishing new models for commerce.
Naturally newer, younger businesses are smaller than the few Web provider
Goliaths. But companies like CD NOW, auto-by-tel, Amazon.com, On-sale
online auctions and many others are starting to produce genuaine success
stories, as are tens of thousands of other small specialty sites. To focus
attention only on Netscape, Yahoo! and Infoseek misses the real action
occurring in the emergence of retail, service, b2b, manufacturing,
consulting and other business segments.
Sean Pfister <seanp_at_cnet.com> suggests:
>readers of this list might be interested in knowing that the ActivMedia
>study is a sample of Yahoo sites, not the Web, and that the study achieved
>only a 16% response rate. With a low response rate, it might be helpful to
>explore non-respondent bias in further research.
Absolutely true. No denials. Low response rate is the bane of direct mail
and direct e-mail research, particularly when conducted among a population
with low investment in study outcomes. The implications are obvious: our
study sample may not reflect the whole web, so projections have lower
reliability than high-response rates of 50% or more. Non-respondent
studies are a good idea for the next round. Are these results reliable?
One might think projections overstated because non-respondent sites might
be less active than respondents. However, we visited all 6700 sample
websites first and know they were actively engaged in Web commerce. In the
other direction, Donna's critique that we have omitted the top 10 sites
suggests our projections may be understated. It's hard to know which
critique is valid, and to some extent both are. One form of validation is
that the business profile of the 1100 respondent websites matches the
original stratified sample and the whole-web population profile (OK, the
Yahoo! proxy) to a high degree. Response rate notwithstanding, the Real
Numbers remains the only systematically collected tracking information of
web-based commerce that is based on the actual information provided by
website managers.
One other point - - projections are not the only meaningful outcome of a
study like ours. A wealth of information lies in the multivariate
relationships from within-website data: how companies create profits under
different business models, how advertising contributes to other web
measures of web success including revenues and profitability, and how the
various forms and styles of website ad publishing compare. One can't draw
this information from inferred numbers, the detail is lacking. For
example, the information suggests that a sizable proportion of all sites
with rate cards have NO AD REVENUES (we say two-thirds, but feel free to
adjust that if you wish) . Websites which use both on-line and off-line
advertising effectively to promote themselves are more likely to be
profitable. Sales revenues are far more important than advertising
revenues for most websites (over one in three have sales, only one in
twenty have actual ad revenues in cash). And we have the ability to break
out sites with advertising / publishing based revenue models from sites
where advertising is supplementary to other revenues.
>ActivMedia claims to sample a known population while
>WebTrack claims to conduct a census, I believe.
Not true. ActivMedia used the widely-representative Yahoo! data as a proxy
for the global web (although blossoming local and regional directories
indicate that that will not be a valid assumption for future tracking
waves) and conducted rigorous n-th name sampling with individual site
visits to qualify respondents. WebTrack is only a census among the sites
in their database (600, I believe) known to advertise and have published
rate cards, far from a census of websites with advertising, let alone a
census of the Web itself, as stated earlier. It is this fundamental
misinterpretation of their research and reporting that perpetuates the
oft-repeated myth that only ad revenues lead to success (and they are not
generally succeeding). Recent articles (Information Week, 8/5, various WSJ
articles) based on this mistaken assertion reinforce the error, misinform
web managers and ultimately lead to misguided strategy and misdirected
investment. Of course the failure of these follow-on sites will also be
repeated as an example of the tricky natures of the web waters.
Thanks for bearing with this response. Let's not quibble over nits - - this
isn't really a debate between studies at all. Its an effort to become
better consumers of web research and better managers of webvertising
resources. Along the way web researchers and marketers should strive to
understand that variety is appropriate, given that the range of website
business strategies, markets, business segments is expanding until it
asymptotically reflects the full range of non-web commerce. The Web is
more than simply a new advertising vehicle - - it is a new marketing
channel.
See you on the 'Net!
Harry
Harold Wolhandler, Senior Research Analyst
hwolhandler_at_activmedia.com
ActivMedia,Inc (800)639-9481 (603)924-9100 fax(603)924-2184
* The Real Numbers behind Net Profits *
See Mid-'96 Web Sales / Profitability figures at: www.activmedia.com
Coming soon:
The Real Story behind Advertising and =91Net Profits
The Real Numbers behind 'Net Profits for Manufacturers
The Real Numbers behind 'Net Profits for Educational Marketers
Harry Wolhandler
ProAction Marketing & Counsel, 350 Water Street, Keene, NH 03431
(603) 357-0721 / fax 355-3355 / hwolhandler_at_top.monad.net
-------------------------------------------------------------------------
This week's Online Advertising Discussion List sponsors:
WEB ADVERTISING '96 How to Bring the World to Your Web Site.
Oct. 31-Nov. 1, New York. $1,000 in Free Web Advertising for each
company sending attendees. Info: http://www.thunderlizard.com/WebAdv.html
-----------------------------------------------------------
>>>>>>>>Announcing the 1996 CyberAtlas Web Advertising Roadmap<<<<<<<<
Need quick answers? Get a copy of the Roadmap and find your way around
click-through, banner design, and much more: http://www.cyberatlas.com
-------------------------------------------------------------------------
