How can marketing research and analytics retain its academic roots and yet be commercially savvy?

I’ve been asked to provide my thoughts on the following question:

“The fundamentals of research are based on statistics, surveys, observation and anthropology derived from the social sciences. As the environment changes, how can we ensure that academic based researchers speak the language of business and marketing? What can we learn from insights companies that have successfully worked with academics so we can best manage this two-way knowledge transfer and which classically ‘non-research skills should be added to the mix?”

I’d like to share my own personal experiences as a practitioner of advanced analytic methods for marketing research and digital marketing for over 30 years; as such, I have tended to work with academics who have very strong modeling and statistics skills. I engaged them to bring advanced methods into play that I was not aware of or would not be able to implement on my own. For example, as Chief Research officer at the NPD Group, I was constantly refining projection and weighting systems to make our data as accurate and as comprehensive as possible. Dr. Richard Columbo (unfortunately prematurely passed on), a proper statistician from the UK Office of Population, Censuses and surveys, who I met when he was a professor at NYU and Fordham University, gave us invaluable insights and suggestions. He also was delighted to have a rich dataset to work with (academics crave data!)

On the other hand, I remember bringing him to a client meeting to discuss some discrepancies between our projections and what the financial services client’s own data showed. By his own admission, his mind was blown (but in a good way!) by this practitioner discussion of issues those discrepancies caused, and practical solutions.

Another example of the interplay of academics and practitioners. One of my roles as a consultant is that I am the Multitouch attribution (MTA) subject matter expert for the Mobile Marketing Association. I have interacted with leading academics to attempt to bring a higher level of understanding to this space.

Briefly, MTA is a newer class of tools borne of digital data where you are able to track many of the ads and content consumption events that a particular user is exposed to and then statistically analyze the patterns vs. business outcomes (i.e. did the user convert to a sale/become a customer.) Working with a number of academics it became clear that assigning credit for a conversion to different advertising tactics is not the same thing as incrementality…(e.g. comparing actual results to the counter-factual of what would have happened if you did not use a particular tactic.)  Just like one has to be careful in any statistical analysis of inferring causality from correlation.

For incrementality, numerous academics gave me the same advice: the gold standard is properly designed experiments…test and matched control, with a treatment variable applied to the test cell. However, an academic cautioned me that the experiment just creates a clean data set; analysis, such as statistically controlling for covariates that might not be perfectly matched, is still required. That, of course, starts blending into practical issues of what variables are important to control for among those data that are available.

However, the purity of experiments is offset by practicality.  MTA gives you an analysis of dozens even hundreds of media tactics/creative asset”/segment targeted” combinations in one analysis. It is not practical to create an experiment plan that can isolate the effects of all of these combinations via design of experiments. Also, there is the human factor.  An experiment becomes a debate…what is the most important media tactic to test? What is the right time of year? And so on.  MTA can analyze media that is running in support of the brand without special experiments requiring running incremental media.

In the case of MTA, interacting with academics brings a north star that I have found to be very valuable. 

However, the goal is business impact and that requires practical compromises.

My position on most marketing problems is a hybrid of the advice from academics and the needs of practitioners. For example, regarding attribution and incrementality, I advise clients to conduct MTA to develop a belief about what works but the proof of the pudding is in the eating…take alternative marketing plans, the current plan vs. one guided by the results of the MTA model, and test them using a proper experimental design. Validation of the set of marketing tactic recommendations from MTA is essential to build confidence in the tool.  For example, marketing mix regression models of weekly or monthly sales and MTA models will provide conflicting findings 80% of the time (based on a survey of marketers) so pure, well-designed experiment bake-offs becomes the only definitive way to resolve the conflict.

Academics bring rigor and advanced methods.  Practitioners have real and robust data sets and real-world problems that academics need to test hypotheses in the context of meaningful marketing issues.

Rigor and practicality, or as Elton John and Bernie Taupin might say, “words and music”…without partnership, it’s likely to be the sound of only one hand clapping.