• by David L. Smith , August 5, 2002

As we approach the reality of reach and frequency programs in the marketplace from companies like Atlas DMT and Nielsen NetRatings, the subject of frequency capping keeps coming up. It has been the assumption on the part of myself and others that this aspect of ad serving will be of great benefit to us as we learn more about what frequency levels are optimum in communicating various strategies.

But once again, everything we know is wrong. After discussing this capability with those knowledgeable in the business, I have come to the conclusion that there is no tool available today which will allow us to control frequency from the campaign side. This is despite the fact that all of the ad servers have talked for years about this capability. It turns out that most of the capability of frequency capping is at the publisher, not the advertiser side. And in order to cap frequency from an advertiser side, you cannot have any more than one ad running at a time! Sometimes when I write articles like this, a solution comes forward, and I hope that is the case this time. But I am not going to hold my breath.

To fully understand this, we need to go back to the basic concept of ad serving. Ad serving was built for sites to organize their inventory and be able to rotate various ads onto pages by request. This enables the sites to sell impressions to an advertiser that could rotate throughout the site, rather than just fix an ad onto a page, reaching the regular visitors to that page over and over.

Frequency capping was first created for the DR advertiser. Analysis of clickthrough rates, and later post-click data revealed that with every succeeding impression, the consumer actions lessened. For many offers, this happened after the first impression for a piece of creative. Typically, once a consumer saw an ad, they were served a cookie indicating exposure to ad “a”. When the consumer was about to be served an ad from the same advertiser again, the server would read the cookie file and serve up ad “b” instead. So, at the rudimentary level, frequency capping is all about creative units, not the overall campaign.

Research by the IAB and other sources has shown that the sweet spot for branding is between 4-7 impressions. So some brand advertisers have attempted to cap their frequency with sites by special arrangement. If they wanted a five frequency, they would arrange for the site to serve an ad from another advertiser after the 5th impression. The problem here is that few advertisers use just one site. And campaigns across sites have duplication. Depending on the number of sites and the paired duplication between the campaigns on the sites, a five frequency on each of a number of sites might be six, seven, eight or even ten frequency overall. So, until we get some more experience with the tools, we will not know. But clearly, there is no way to control frequency for the campaign through individual site controlled methods.

Enter the third party ad server. From a third party standpoint, I should be able to know what the frequency is cross-site and solve this problem. Except for one detail. The third party ad servers cannot dictate to the site what advertiser will be requested to display the next ad. A third party server cannot tell the site not to serve up an ad from an advertiser if the site has an order from that advertiser to run x number of impressions. As such, all of the major third party ad servers that are deployed for agencies and advertisers can only cap creative, not a full campaign, taking us back to where we started. The result is that agencies who are telling clients that they are capping frequency are really only doing so by individual creative. If they have five or even 10 pieces of creative (not unusual), they could be serving 20 or more impressions to a single user whom they are “capping” at five impressions.

We need a technology solution. It does seem that a company like DoubleClick, who is the leading agency/advertiser solution and one of the top two or three site side serving solutions, could deal with this problem by having DFP and DFA communicate with each other. Or maybe someone else has a workaround that has not been defined.

In the end, we need a solution. Won’t do much good to deploy reach and frequency if we can’t act on the learning. The Interactive medium has the theoretical controls. Now we need to make them work in the real world.