According to the
NY Times: “A group of large companies, including
Kimberly-Clark,
Colgate-Palmolive and
Ford Motor have said that by the middle of 2007, they will demand that online publishers hire auditors to check their ad and viewer counts.”
As they should. But perhaps amidst the brouhaha, someone can explain me the following:
I have always been very suspicious of online panels that measure audiences. Both Media Metrix Comscore and Nielsen Net Ratings use panels.
I think it’s a valid, noble attempt to measure traffic on websites but it has its shortcomings.
This is what I think would work better: a site-specific measurement that accounts for all of the traffic on all websites. In a nutshell, think of taking Google Analytics’ stats, putting them on all sites and compiling results of the largest websites using every site’s traffic stats from the same source. It would be pretty interesting to see traffic ranks when you compare apples with apples and oranges and oranges. For more on my thoughts here, click
here.
Of course, there is the issue of cookies, here’s what a venture capitalist who backed Comscore had to
say about the matter:
Comscore counts real viewers in its panel, not cookies. Cookies get deleted by spyware removal software. If you remove your cookies once a week, you’ll look like four users every month to someone using cookies as a basis for UVs. The more sophisicated a user base is, the more likely they use cookie removal. And that results in significant UV overcounting.
Notice the major assumption there: “If you remove your cookies once a week.” I do not remove my cookies all that often and I’d like to think I am somewhat web-savvy. And, I probably use spyware removal software less than once a month. I’d be really curious to know what the average on that is amongst web users, and at the risk of being facetious, if I did want to venture and find out, I know I wouldn’t use a panel!
Alas, I am not here to argue about panels. I just want anyone to explain to me the following:
In the context of
a) Advertisers asking publishers for audits
b) Comscore and Nielsen Net Ratings usually being a mile apart
c) Most sites looking at internal or third party log files
d) Advertisers using their own, third party audited ad servers
e) Most websites using their own, third party audited ad servers
How could anyone defend panels when advertisers themselves spend ad budgets in a manner that suggests that panels are way off and erroneous?
The agency “added value” actions can be broken up into planning, buying and creating.
I will be the first to argue that Comscore and Nielsen Net Ratings are super important in the planning stage. This is driven not only by Comscore and Nielsen Net Ratings’s core products but also by their psychodemographic products AiM and @plan (forget my thoughts on these…).
At the buying stage, it’s less important (more on this below). Comscore and Nielsen Net Ratings help media buyers and planners justify their decision and defend themselves if need be but at the buying stage, they are almost forgotten!
Say Comscore and Nielsen Net Ratings give a site 1-2M uniques and 5-10M pageviews. For the sake of this example, we’ll go with one source that gives the site 1M uniques and 5M pageviews. Speaking from experience, one can see that according to both this panel’s numbers and the site’s internal/third party stats, users generate 5 pageviews per unique user.
Yet, when the agency wants to buy traffic, they ask the publisher for an estimate of impressions, share of voice etc. and expect that a website will use their own site’s numbers.
In my experience closing over 1,000 ad deals in 5 years (probably more), not once did an agency say “but wait, these pageviews you are selling me are mugh higher than what Comscore/NNR is crediting you with.”
At my old employer, panels gave us numbers in the range outlined though our log files gave us 6M unique users and 40M pageviews, roughly 6.66 pageviews per uniques (fitting, perhaps but that’s another story). Anyway, when a buyer would ask me occasionally for a site-wide creative (say a text link on all pages), I’d tell them they “should expect 40M ad impressions (with one ad impression per page).”
When the campaign would finish, using our ad server and cross-referencing it with the advertiser’s ad server (usually aQuantive’s Atlas or Doubleclick’s DART), the end number was in fact very close or equal to 40M impressions, suggesting that indeed, our site got 40M pageviews.
Yet, Comscore/NNR would give us 5-10M pageviews. How come? That’s three sources saying 40M pageviews yet Comscore/NNR come out and shortchange us with 5-10M pageviews.
Even the advertiser who spends money concedes that Comscore/NNR are wrong, no?
So pro-panellers: we’re not talking [new] unique users vs. [repeat] visitors. We’re talking pageviews. How can anyone explain such a high variance in pageviews between:
a) a site’s third party log files AND a site’s ad server AND an advertiser’s ad server,
versus
b) Comscore and Nielsen Net Ratings.
I am opening this up for debate. Anyone?
Subscribe:
November 7th, 2006 at 10:34 am
[…] Cookies and pageviews not requested by or seen by end users have become the boogieman, or dare I say the cookie monster in the industry. Cookies or unrequested/unseen pageviews could explain 10, 20, 35% discrepancy but when your numbers are 50-75% off, how much cookie can a man have? Furthermore, we’ve already explained here that this argument is a red herring. Truth is that if this was the case, advertisers would never pay what they buy. Advertisers might plan media buys using panels’ stats but they buy using a site’s numbers. Read more. Regarding the article’s assertion that online advertising is a crapshoot, we categorically reject the statement as ludicrous. […]