• by David L. Smith , October 19, 2007

Attribution has always been an issue for ROI-based Internet campaigns. Historically, ad servers have been programmed to credit attribution for a sale, download or other action to the last click from the IP address. This works to the advantage of the search engines and may, in fact, be a primary factor in the high valuation of Google.

The system works like this. When a campaign uses a third-party ad server (3PAS), the computer is cookied when an ad is served to the browser. Then, whether or not there is a direct click to the directed site or whether the consumer goes there later, the cookies is referenced and attribution given to the last site and ad the consumer was exposed to. When sites or networks serve the ads, each one of them can claim attribution if they "see" the consumer. That's why 3PAS makes sense, to dedupe and give credit to the true last view or click. Historically, the networks, especially when a CPA deal was involved, served their own ads rather than have them served by the agency 3PAS. This presented problems, as the networks would try to set things up so that they saw the maximum number of uniques and could claim the maximum bounty for those visiting the clients' site. Attribution claims under this kind of setup would sometimes exceed the total volume garnered by the client. Enter the Universal Action Tag created by Atlas and a similar product from DoubleClick that enabled all ads to be served from the same server.

But the last-click problem continued. Here's an example of why last click as sole attribution is broken: This is from a case study example provided by dgm at the iMedia Agency Summit in Hunter Valley Australia in September. The client had a search effort garnering 3,000 customers a month at a $40 CPA. They had 1,000 customers from the Web at a $130 CPA. Other efforts like affiliate and email had a $60-90 CPA, garnering 2300 customers a month. The client directed the agency to cut the "expensive" Web CPA sites. Agency did so, getting it down to $100. The 1,000 customers dropped to 650. The 2,300 dropped to 2,080. The 3,000 in search dropped to 2,250. Overall volume dropped by 21%.

Net, net: Web banners, even though they have a higher CPA, enable the overall scaling of a campaign to higher volume. When they took away key sites that contributed to the branding of the advertising, fewer people knew about the offer, and fewer went to search.

Enter companies like Blackfoot and Theorem and their tests with what Publicis, Theorem and we at Mediasmith call Multiple Attribution Protocol. These tests, run last year, resulted in agency demand for Atlas and DoubleClick to push up product development and provide reporting of ALL impressions served to a consumer and all actions taken before the last action. DoubleClick is currently in beta with this service, which the company calls Exposure to Conversion reporting. Atlas expects to release its product, called Advanced Conversion Attribution, in Q1. At OMMA last month, we had a lively panel on this topic that included Dave Falk from DC, Scott Baker from Atlas and Jay Kulkarni from Theorem. The news is encouraging.

From an industry standpoint, we look forward to the maturation of these products and work toward the algorithms that will help us assign credit more properly across all of the touchpoints of a campaign. While this may take some of the bloom off search, it should be a boon to all digital advertising, as clients are able to more readily identify actual consumer behavior prior to purchase. These concepts can then be transferred over to other media as they become digitally enabled.

Farewell, last click. Hello, MAP. And welcome. We have been waiting for you. Now go to work.