Happy Easter/ Good Friday!
I read a Google Analytics paper (click here) recently about digital attribution and how most advertisers are unhappy about last-click attribution but yet 54% continue to use last-click attribution.
For the uninitiated, attribution is the process of assigning credit for a favorable outcome (e.g., a sale, click on a request for more information, registration to a mailing list, etc) on a website to the various sources that led the visitor to the website. The complication arises because a visitor may click on a search ad on her first visit, a banner on her second visit, and eventually buy during a visit initiated by an organic search. How would you attribute the sale to these different sources? Last-click attribution is a technique where the last source leading to the favorable outcome gets all the credit. Most people agree that's a dumb way of looking at the digital/multi-channel/ multi-device world (but are doing precious little about it).
Attribution is important because it allows the advertiser to correctly price the value of a keyword (on a particular search engine) or the value of TRPs on a network TV ad or the value of a banner on a particular social media website or the value of an affiliate lead. The accuracy of the value computation will become a source of significant competitive advantage in the emerging digital and multi-channel era.
Some smart folks have moved beyond last-click attribution to give weightage to sources contributing to visits leading up to the favorable outcome. My question is : How do you know that your attribution technique is working?
My suggestion is that the advertiser should test various attribution techniques to understand the pricing of various keyword-engine, banner-publisher, TV TRP-creative combinations using data over recent time periods using these alternative techniques. the technique that yields a statistically stable estimate of pricing should be chosen.
The other thing that an advertiser should do is analyze the contribution (to a favorable outcome) of a first visit, the last visit, a visit in the last n days, etc by type of source (Search Engine, Publisher, Social media website, affiliate, etc) using a decision tree. This will help understand typical patterns of visits leading up to the 'successful' visit.
