This Half Is Wasted

You may know the expression, “I know half my advertising is wasted, I just don’t know which half.”

Well, I know which half.

My group is constantly asked, “What is this (fill in the blank) worth?” Sometimes, the context is an integrated online/offline program, and we’re asked to evaluate the fair market value of the online element. Other times, it’s just an online submission as part of a request for proposal (RFP) process. We must place a value on the proposed package.

Our typical methodology is first to break down the proposal elements into its various components (not always an easy task). Once we have a concrete idea of what elements we’re getting, we apply a little, secret formula. The formula falls under the if-I-told-you-I’d-have-to-kill-you category, but I’ll give you the gist: composition of said URL’s target audience x number of impressions x ad impact factor x subjective immersion rating.

Let’s focus on the first element, audience composition. One amazing thing we routinely find is the relatively low target composition of some of the most targeted Web properties. Though it’s true syndicated research isn’t granular enough to provide demographic breakdowns of all channels and subchannels of a particular site, according to comScore Media Metrix in July 2003 only 65.4 percent of iVillage’s audience was female; 21.7 percent of’s was 12-17 (if you add 18-24 year olds, you only get 38 percent 12-24); 59.4 percent of’s was male; and only 49.2 percent of’s were users with a household income of $75,000-plus.

Can you believe that?

I can. After doing these types of analyses for many years, I can safely say the most targeted Web properties, demographically speaking, will only deliver 50-60 percent composition of your target. Right off the bat (before applying our formula), 40-50 percent of your impressions are wasted inasmuch as they don’t reach the intended target.

Why don’t we buy the Internet on a demographic basis? Traditional media (especially TV) use demo CPMs to understand delivery against the intended target. If any medium were to welcome the ability to micromarket and maximize audience delivery, wouldn’t it be the Internet?

Reach and frequency projections from syndicated research do not correlate at all with actual serverside (cookie-based) campaign delivery.

This is a serious problem directly related to the above issue. Fact is, in three separate instances in which we compared the two, the reach and frequency of a campaign that entered into a syndicated research tool differed from actual DoubleClick delivery data from -81 to 90 percent. I’m no statistician, but that’s an unacceptable fluctuation.

One could speculate the huge discrepancies between the panel data and the serverside cookie data are due to a problem with the panel sample sizes or the lack of resolution between the work and home audiences. It could be a methodology issue. It’s my belief the issue primarily is rooted in the lack of specificity with which one enters a buy into a syndicated reach/frequency (R/F) system.

Interactive buys are often a complex mix of home page impressions, run of site (ROS) impressions, channel impressions, subchannel activity, and many other permutations. Syndicated research companies report on top-level domains and selected primary channels. That’s like comparing a ballpoint pen to a paint roller.

What can we do? We can start making interactive plans and buys more codified than the wild variability that exists today. We can try to more closely match the way we buy media with the way it’s reported to reduce the level of ambiguity. We may even be able to add some sort of “soft” measurement that would allow planners to massage R/F outputs based on past campaigns or a particular buying style. Something must be done. We cannot continue to report reach and frequency numbers not grounded in reality.

I call upon interactive marketers to lend a hand. Let’s stand together as a community to change the media planning and buying process. Buying interactive media on a demo CPM basis and having accurate R/F projections would be two really good things for our industry.

What do you think?

Related reading