Hurdles in Attribution.

Recently, CIMM, The Coalition for Innovative Media Measurement, released the findings of a study on attribution conducted by Sequent Partners. The main conclusion was that, as an industry, we have made some progress towards the development of an attribution model across media. However, there is more work to do and, at least in the near term, there are few viable systems currently available that holistically capture cross-platform consumer behaviors.

This is an industry imperative but the challenges are steep. To get a lay of the land, I sat down with Julian Zilberbrand, EVP, Audience Science, Viacom Media Networks, Ben Clarke, President, The Shipyard and Jane Clarke,  CEO, Managing Director, CIMM.

Charlene Weisler:  What, for you, were the most important takeaways from the CIMM Sequent Whitepaper on attribution?

Julian Zilberbrand: There is still much work to be done to truly do cross platform attribution, especially when taking into account non-digital touch points.  The methodology used, I believe, still leads back to a Marketing Mix Modeling (MMM) versus Multi-Touch Attribution (MTA).   MMM has its value but is a different capability that has been in existence for a long time and is seen to have flaws as well. The need for attribution is real and necessary in our space but individual advertiser KPIs make a one-size-fits-all model impossible.

Jane Clarke: The most significant insight emerging from our whitepaper is that clearly while progress is being made, there is still much development work needed for a truly integrated marketing mix modeling and attribution approach to ROI analysis. We have outlined some important elements of the direction that development needs to take for a more holistic solution, but important work lies ahead.  The future of cross-channel attribution is being formed now, so it is in the best interest of media buyers and sellers alike to get involved in the process to ensure that the industry has options moving forward that truly serve our needs.

Charlene Weisler:  What do you see as the greatest challenges in attribution? 

Julian Zilberbrand: The biggest hurdles from the digital end are the proper attribution of search and social; you need actual log files that search and social partners will not provide. As for non-digital touch points, there is a lack of universal id to properly attribute exposure in more traditional areas of marketing and the proper data structure with functional nomenclature across all channels.

Ben Clarke: There is so much noise in the data as to what's really influencing a sale - something as simple as how close a person lives to a retail location is a main driver of sales - but not accounted for in attribution - further most attribution models.  In addition - the term attribution itself probably needs to be dimensionalized - attribution models are usually deployed against sales targets - but some channels may be great for building awareness and preference (which are important) but don't lead directly to sales.  In other words - attribution deals in Purchase Funnels where not every meaningful relationship is a sale.

Jane Clarke: The biggest challenge to full cross channel attribution is accurate measurement of consumer and/or household identity to link all the media exposure datasets to purchasing datasets and/ or other KPIs for marketing campaigns.

Charlene Weisler:  How long do you think it will take for formulate a viable attribution model? Is it even possible? 

Julian Zilberbrand: We still have work to do to get to MTA to be a truly viable solution for determining effective marketing touchpoints across all channels.  Not all consumer touch points are given proper credit or provide the level of time-stamped data which can effectively feed the MTA properly. Additionally, various advertiser KPIs present a real challenge to solve for in a single modeled approach. The model might not yield relevant results in every instance.  Creating models or game theory takes a significant time to ramp up and requires a level of patience that most marketers don’t have.  Of course it’s possible, but only if cross platform measurement solutions, both globally and for the U.S., reach their full potential.

Ben Clarke: I don't think there will be "a model."  This is the issue with one size fits all fractional attribution platforms you can buy out of the box - the neatly assign values to each channel - but when you go to activate against those insights they largely fail to be predictive.  Usually great attribution involves setting up controlled experiments on a continuous basis - so I don't think the goal should be to come up with "a model" at all - maybe better to come up with a methodology that can deliver versions of models over time.

Jane Clarke: Viable models are beginning to emerge for full cross channel marketing ROI attribution, and will improve along with the ability to link and/or model consumer and household identity across more and more devices, marketing channels, and purchasing datasets.

This article first appeared in

No comments:

Post a Comment