• by John Kelly , June 10, 2011

When a sale comes in, almost every marketing channel raises its hand to claim credit. But do they all deserve credit?

A growing number of companies -- Convertro, Marin, Tagman, Turn -- attempt to answer this question for marketers. With that comes a wide variety of attribution models. Which one is best for you?

The best place to start is to determine what the various models can and can't do. To illustrate the differences, let's take an example of an online shopper, Imelda, who is buying shoes. She gets a catalog in the mail and sees a few pairs she likes, types in the company name to Google and clicks on an SEM link. She shops around, leaves the site, sees a display ad, comes back to the site a day later, finds an affiliate coupon code and buys the shoes.

First Touch Attribution - This model puts the entire emphasis on the marketing channel that first brought the user to the Web site -- it online or offline (catalog, TV ad, or radio). Credited with Imelda's sale: Catalog.

*Advantage -- it takes into account offline marketing efforts as well as online.

*Disadvantage -- it disregards many of the marketing channels that convert a visitor to a buyer.

First Click Attribution -- Like the first touch attribution model, this model gives credit to the online marketing channel that first drove the user to click to your site. Credited with sale: SEM.

*Advantage -- it recognizes the importance of a new user.

*Disadvantage -- it ignores the marketing channels that convert visitors to buyers as well.

Last Click Attribution -- This model takes the exact opposite position from the first click attribution; it gives full sale credit to the last marketing channel that the user used to come to your site before making a purchase. Credited with sale: Affiliate.

*Advantage -- it emphasizes the marketing channel that is closest to the actual sale.

*Disadvantage - it fails to consider the marketing channels that drove the user's interest in the first place.

No Click Attribution -- This model puts together a model of your marketing spend across various display ads and draws conclusions about which parts of your campaign were most influential in driving the user to the sale. Credited with sale: Display Ads.

*Advantage -- it takes a concerted effort to map the beneficial influence of online display campaigns.

*Disadvantage -- it does not consider those marketing channels which are directly driving people to your site.

Multivariate Attribution -- This approach gives each online marketing channel that a user is exposed to a certain percentage allocation for the sale. Sometimes those percentages are preset (i.e., each marketing channel has equal weight); sometimes those percentages are configurable. Credited with sale: SEM (33%), Display Ads (33%) and Affiliate (33%).

*Advantage -- it takes into consideration all of your online marketing channels.

*Disadvantage -- The percentage allocations are generally static and not weighted by user experience or product categories. Essentially, if users experience the same marketing channels, then they are assumed to respond the same. Further, since these models are only as good as the allocated percentages, and any improper percentage allocation will skew results.

Customized Attribution -- This catch-all essentially covers all of the other attribution models specifically developed and customized for retailers by analytics companies. Credited with sale: TBD.

*Advantage -- You can get a good picture of your user's online click behavior through a customized model.

*Disadvantage -- It can be expensive and time-consuming to establish and even more difficult and time-consuming to perfect.

There is significant disparity in the models for two reasons: First, modeling is in its infancy, and second, many channels can influence an online purchase. Experiment with models and marketing mixes until you start to see patterns develop. The patterns will tell you what is working.