Written by Brook Schaaf

You surely know the famous Wanamaker quote: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” Some point out that Wanamaker may never have said this, but, apropos of the subject, it still attributed to him, which makes it even more savory that it might be ironic.

As a young affiliate marketer, I seized upon and frequently restated this sentiment as a sort of proof of the excellence of affiliate marketing, which was then uncontroversial — at least the last-click part of things. 

Fast forward to today, and there is much gnashing of teeth and rending of garments over the faultiness of assigning credit to the final touchpoint because there may have been prior touchpoints, and the last one may not deserve credit for the conversion. First-click is the same problem in reverse — the negation of more influential subsequent clicks, assuming the click deserves credit to begin with. Models like Linear, decay, weighted, U-, or W- shaped attribution attempt to split the baby in different ways.

Moreover, all assume a click as a signal when the user may bypass the click to a conversion, for example, by changing devices. But what else can you do? An attributable signal may not be visible to the advertiser. 

This is not to say that the criticisms (misallocation of ad spend, favoritism toward bottom-of-funnel, and reduction in the potential of certain partnerships) are unjust, but that they should be considered against other options and their shortcomings. What it boils down to is that nothing is perfect, hence the joke that attribution is ultimately an article of faith.

Last (or other) click attribution is not unique to the affiliate channel. If you run, say, multiple keywords in AdWords, the last one usually gets credit unless you change to first-click or a split credit. And the click is likely to be navigational in nature, as with a branded term. If I see a billboard for, say, the Austin rodeo online, then click on a paid link instead of an organic listing (remember those?), what value has Google contributed? Probably some, but not as much as with a keyword that lets me discover the rodeo for the first time.

Mix Media Modeling (MMM) has been making a comeback, but you should ask why it lost favor to begin with: time (years of data); cost (both high ad spend and high analytical costs); lack of granularity; and, frequently; lack of actionable data.  It has always struggled in the fast-moving, dynamic, and trackable digital world, hence the adoption of views and clicks.

While software improvements may make MMM more affordable and useful, it seems more likely to me that marketing managers will just get overwhelmed by the data — and affiliate will most likely get the short end of the stick. You may call me cynical, but I think Google’s push for MMM is a tacit admission that signal loss is a growing problem, and it suits them just fine to insert a refangled probabilistic model if they can no longer proffer a deterministic one. 

No attribution method is perfect, and none ever shall be. We might then re-tooth Wanamaker’s old saw: “Half my attribution is accurate; the trouble is I don’t know which half.”

The post Attribution Considerations appeared first on FMTC.