In my last Analyze This post, I alluded to the importance of click to conversion time, that is, the estimated conversion time of the revenue attribution window. I shall devote this article to explaining the importance of the revenue attribution window and some empirical methods to help determine the right window length.
The revenue attribution window refers to the maximum length of time between click and conversion that an advertiser must include when calculating the revenue from a click. Consider two consumers who clicked on an ad for a t-shirt on a given day. One consumer clicked on the ad and bought the shirt on the same day. The other consumer did not convert the same day but bought a shirt from the same website two months later.
Intuitively, we know that the ad was instrumental in the first conversion but perhaps had a very minimal or even no role in the second conversion. If my revenue attribution window was 10 days, I would consider the first purchase but not the second one when I calculated revenue for the keyword for that day, but if my attribution window was 60 days I would include both purchases. Clearly, the attribution window can have a big effect on how profitable a campaign or keyword appears.
So how we set the right attribution window? There are several complex ways to calculate this using methods such as gamma functions, hazard modeling and so on, but I shall present a very simple heuristic using conversion time that will get you started.
The following graph analyzes the click to conversion times of 2500 purchases on a retailers website for a clothing campaign.