Ad Metrics: Where the Rubber Hits the Road

How does your client know that all that money shes blowing out the media chute actually has some sort of positive effect on her business? In traditional media, she looks at several numbers to either determine a before-and-after effect, or just to guarantee a certain amount of exposure. These are the numbers agencies have typically been responsible for collecting and presenting as a post-buy document.

Traditional Performance Metrics

Its worth going over some of the terms they use for this purpose. Well see them pop up frequently in our online dealings. (Important terms are italicized in this series.)

Advertisers tend to break up audiences into groups of folks or markets, if you will. In a wonderful demonstration of our industry myopia, we call the particular market segment were thinking about a universe.

The two major media measures of the universe are the share of voice (SOV) and the share of market (SOM). The share of voice represents the percentage of impressions your ads comprise within your competitive categorys media spending against a particular audience universe. The share of market represents the percentage of use your clients product receives versus the competition within the audience universe.

This is normally where an explanation of ratings, ratings points, GRPs, TRPs, and the like would come in. Ratings are one of the measures used to determine SOV in broadcast media. Im going to save that for a coming column, however, so we can spend the proper amount of time making fun of it. To tide you over, Tom Hespos intelligently approached the issue last week.

We all figure that the greater our SOV, the greater our SOM. But it doesnt always work that way, and you can usually bet your next expense check that the company that sees its share of market erode well past its share of voice will fire the agency in the next couple of quarters. I know a guy who actually trades the stocks of agencies specifically with this in mind. And he doesnt do too badly.

Recall is the measure of the percentage of folks who can remember seeing your ad. We usually speak in terms of either unaided recall or aided recall.

Unaided recall goes like this:
Q: “Do you remember seeing any recent beer commercials?”
A: “Er, I dunno.”

And aided recall goes more like this:
Q: “Waasup?”
A: “Waaaaaasuuuuuuuuup!”

This recall of the creative hopefully leads to an increase in preference. Technically speaking, preference is a ratio (expressed as a percentage) of the rate of consumer choice of your client’s product versus the rate of choice of others. This is truly where the rubber hits the road. You found the target, you got a media message to the target, the creative had a discernable effect on the target. Agencies increasing the preference percentages for their clients brands seldom need worry about being fired.

The Online Advantages

In the online world, we have quite a different set of performance information. In fact, we often have a complete record of everyones interactions with our ad. We know who saw it, who clicked on it, how much stuff they bought, and what profit margin they represent going forward. The trick isnt so much getting this information, as it is linking one event to the next, resolving discrepancies among data sources and engineering a process to digest this information and exploit it usefully in our media context. These are all key areas we shall explore in future columns.

I want to stress that solving these problems are much more important than attempting to replicate our traditional media performance metrics on the Internet. I see a lot of traditional agencies wasting a lot of time worrying about things like share of voice within a site. That energy would be much better wasted worrying about the return given to the client relative to the cost of the media investment especially since those numbers are there to be ferreted out by a competent media buyer.

Get your machetes out, because next week were going to go through the jungle of online metric vines and thistles. In the process of explaining how we can link the observed behaviors online, well cover banner servers other technologies that never seem to give us straight answers. Well do some explaining about the technical causes of this, hopefully setting up your expectations as to what can be usefully garnered from the various online measurements.

Related reading