Optimizing On Cookie Crumbs
I strongly believe that if we ignore consumer behavior (you might want to read my last post on landing page testing), we are doing ourselves and our clients a disservice. It's very possible that each exposure is necessary in driving the major influence for the conversion, yet we typically optimize based on last click attribution anyway. In the case of online media, where we can measure the influence, why aren't we? In contrast to offline media, I don't think any client is willing to risk turning TV off, even if we can't directly quantify and attribute the results to a specific commercial or time slot.
To test the online multi-click attribution theory, we developed a custom report based on campaign exposure conversion data (i.e., the exposures or clicks the consumer had leading up to the conversion), to determine if conversions occur after the first initial ad exposure, or if the results are being influenced by the synergy of multiple sites. Think of it as counting the crumbs (or impressions) you leave behind you when you take the cookie from the jar.
Looking at a recent one-month campaign in May 2010 for one of our finance clients, almost 60% of conversions occurred after the exposure to one ad, and the remaining were the results of 2+ exposures. As expected, search engines played an important role in driving direct conversions. Based on the last click/view attribution model, both Google and Yahoo generated almost 55% of the conversions. Individually, the display sites on the buy don't look like strong conversion drivers.
Taking a closer look at the campaign results using all of the impressions and clicks prior to the conversion, the display buy influenced 70% of all campaign conversions. AOL generated close to 15% of direct conversions, but almost 20% of all the people who converted were exposed to an ad on AOL before converting. Casale also proved to be a site that was efficient in generating awareness (third highest percentage of total exposures prior to converting).
While this is only one example on a relatively small data set, multiclick attribution provides a more complete picture than last-click or last-impression attribution. If a consumer was influenced by a series of online ads prior his conversion, each site is a necessary and influential player in the process and we shouldn't be optimizing out of them if they don't provide a direct conversion. Adding consumer value (or size of the cookie) to each of the conversions to calculate ROI tells a whole different story (but that's for another time).