Commentary

Accelerating Attribution

Attribution, according to the Coalition for Innovative Media Measurement (CIMM) Lexicon, is “the reason that a prospective customer does what they do. More specifically, why they came to the site, entered the funnel, and performed a given action. (Source: Mediamath)”

Accurate attribution would enable a marketer to get the correctly measured sequencing of consumer-initiated events that influenced that consumer’s behavior. It is not an easy task.

CIMM is hosting the Attribution Accelerator Conference, to be held today, presented by Sequent Partners, Time Inc., Viant, and GreenBook. Speakers from such companies as Turner, Nielsen, comScore, Citibank,  Rubicon, Google and Verizon will present their views on attribution.

Questions to be addressed include: Where is attribution research today? How can we spur development to ensure existing models can keep up with demand as attribution moves beyond digital?

I sat down with Jim Spaeth, and Alice Sylvester, both partners at Sequent Partners, to discuss the topic further.

advertisement

advertisement

Charlene Weisler:  What are the best forms of attribution?

Alice Sylvester: Attribution is the study of individual media tactics’ contribution to sales. The best forms of attribution do not use a priori models that presuppose the solution (first click, last click, etc.).  The best forms of attribution incorporate other elements besides digital, at the household level.

Weisler:  Tell me what your company is doing to track and measure attribution.

Sylvester: We have a number of projects going on — notably the Attribution Accelerator event, which will quicken the pace of innovation, fortify the science and galvanize the industry toward new attribution solutions. We also have a CIMM study of attribution that will assess best practices and areas of improvement in cross-platform attribution.

Weisler:  Which existing models are doing the best job of attribution?

Jim Spaeth: We don’t know. We will be looking to identify a best practice over the course of the next few days. We do know, however, that many of the existing models are insufficient, given that they need to: reflect the fact that media has diminishing returns…, that media has short- and long-term effects…[that there are] different responses by consumer segments and past purchasing behavior, and that different ads generate different responses.

Sylvester: The best job of attribution will address these issues.  Oh, and then there is the issue of walled gardens, and a significant portion of data from Google and Facebook not being generally available.

Weisler:  How can you ensure that existing models keep up and move beyond digital?

Sylvester: The marketplace will definitely determine winners and losers. Skepticism is creeping in among marketers who have had some experience with attribution.  

Spaeth: There is a huge issue on how to impute non-digital events on a digital data stream.

Sylvester: There are issues with validation and the veracity of A/B tests that have to be addressed in order to increase confidence in attribution.  And there are significant issues in device matching at the household, which can be a source of significant error, and in data matching across big datasets.

2 comments about "Accelerating Attribution ".
Check to receive email when comments are posted.
  1. Ed Papazian from Media Dynamics Inc, November 30, 2016 at 9:24 a.m.

    Some interesting comments by Jim and Alice. As they point out, attribution---when appropriate----is not an easy thing to measure, especially as ad campaigns wear out over time due to their inherent redundency. It's not just a matter of the consumer entering "the funnel", then being exposed to one or more ads for the brand and, finally, buying it. In fact, for established brands something like 75-80% of the measurable sales effects are nothing more than already sold consumers restocking their supply of the product.

  2. Charlene Weisler from Writer, Media Consultant: WeislerMedia.blogspot.com, November 30, 2016 at 2:15 p.m.

    Hi Ed, Thank you for your comment. I agree. But maybe with machine learning and more programmatic applications we can learn more abut what triggers an action. This will take time to fully explore though.

Next story loading loading..