Marzo 31, 2010

The Varying Impact of Ad Frequency in the Digital Environment

Di: Gian Fulgoni

Last week at the annual ARF Re:Think conference in New York, Hernan Lopez, President of .Fox Networks, and I presented an analysis of four ad campaigns run on the .Fox Network in the U.K that were evaluated using the comScore AdEffx™ platform. One of the campaigns used only video ads, one used a combination of video and display, while two campaigns featured display ads only. The four campaigns delivered a total of 300 million impressions. We used the comScore panel of about 80,000 U.K. Internet users to understand how display ads and video ads increased visitation to the advertised brands’ Web sites and how they increased search queries that used trademark terms or related generic terms. For each campaign, our analytical approach compared the behavior of panelists who were exposed to at least one ad with the behavior of a balanced “control” group of panelists who were not exposed to the ad campaign.

Despite a minimal number of clicks on the ads, we found that each of the campaigns successfully lifted site visitation and search queries, revealing both the latent branding impact of online ads as well as the inability of clicks to accurately reflect campaign effectiveness:

For the one campaign that used both video and display, we were able to directly compare the relative effectiveness of the two formats, which showed that, while both video and display successfully lifted site visitation and search, video was more effective:

We were also interested in understanding the way in which response to frequency of exposure differed between display and video ads. To do this, we compared the behavioral response across groups of panelists who were exposed to varying numbers of video and display ad impressions. Below are the results when we examined the lift in site visitation:

What’s very interesting is that the lift in site visitation following exposure to display ads climbs steadily with increased number of exposures, while the lift following the video ad exposure occurs relatively quickly (within 1 to 5 exposures) but then increases only slightly with additional exposures beyond 10 or more.

This frequency response pattern is illustrated even more vividly when looking at the lift in search queries:

We can see that the lift in search queries following exposure to a video ad occurs strongly within the first five exposures but does not increase with additional exposures. In contrast, the impact of display ads continues to climb markedly with exposures beyond the first five.

What are we to make of these results? One implication is that display ads are easier for consumers to miss than video ads. As a result, more frequency is needed to gain consumers’ attention and engagement with display ads than with video. Hence, the steadily growing behavioral response with additional exposures of display ads.

If this supposition is correct, then there is a potentially positive implication for advertisers and publishers. Since cookie deletion renders an ad server incapable of distinguishing between a true unexposed browser and a browser that has already received an ad but has had its cookies deleted, ad servers will repeatedly deliver ads to already-exposed browsers. As a result, many display ad campaigns actually end up delivering a higher-than-planned frequency (i.e. a higher-than-planned number of display ad exposures per browser). While the conventional wisdom says that such over-delivery results in wasted impressions, the results of our study suggest they may not be wasted at all since it appears that a high number of repeated exposures are needed to gain consumers’ attention and engagement and to elicit the desired change in behavior.

I do want to stress that the conclusions I’ve drawn here are obviously based on a limited number of U.K. studies and as such may not be broadly representative of all display ad campaigns running in the U.S. or the U.K. Nonetheless, they do suggest a potentially valuable avenue for future research, namely identifying the sales lift obtained from varying ad frequency levels. To that end, we would welcome hearing from advertisers, publishers, ad networks and agencies that would like to participate in a broad study of U.S. online ad campaigns aimed at measuring sales response (both online and offline) as a function of frequency of ad exposure.

Looking forward to hearing from you.

Tag: Advertising