Descriptive Research and Insights are not Enough

Broadly speaking, marketing research studies fall into two classes…descriptive and predictive.  Descriptive research includes things like segmentation, A&Us, qual, even brand trackers which are retrospective in nature.

One of the biggest challenges to marketing research is when actions from the insights are not clearly indicated to marketing.  The reason? We researchers do not try hard enough to flesh out the predictions embedded in the insights by creating a math structure to our findings.

Here is a negative example: Typically, we analyze tracking data and find that a brand is not rated particularly highly on an attribute that is highly correlated to brand preference.  So, in our presentation, we stress the importance of improving that attribute rating.  But how? Telling creative teams to do better?  Is that attribute even movable?  For example, if you apply a math structure to attribute ratings, you will realize that attribute associations that are really low are also really hard to move. You are better off finding attributes in a mid-range of ratings that are also correlated with preference. Those are easier to move with advertising.

Here’s another negative example: I tested the sales potential of a new product where we included questions needed to classify respondents into segments that an innovation consultancy had delivered to the client that led to the new product idea. The segmentation made a lot of intuitive sense but guess what? The consumers in the segment that motivated the new product idea did NOT have any higher purchase interest! Clearly, the segmentation was useless but that was only revealed by examining its veracity by testing the implied predictions.

Now, take a look at a positive example: I have always known that you can model the distribution of consumers in terms of their probability of purchasing the brand of interest using a Beta distribution.  OK, that is descriptive…where is the prediction? So, working with the MMA and Neustar, and fueled with Numerator data, using agent-based modeling and calculus, we discovered that those in the middle of the curve…those we called “Movable Middles”…were mathematically expected to be most responsive to advertising for the brand.

Across a dozen or so cases, this math-driven principle has been proven to work 100% of the time (what else in marketing offers such a guarantee?) Most recently I consulted with Viant, a DSP to design a test of Movable Middle theory with Circana (fka IRI) frequent shopper data. We found for three CPG campaigns that the average lift in sales for Movable Middles was 14 times higher than those not in the Movable Middle. This is how you take a descriptive model (Beta distribution) and find the prediction value and actionability (push a list of IDs in the Movable Middle for programmatic activation).

About 5 years ago, I made two predictions.  I predicted that Amazon would become the number 3 media company in ad revenues and that Netflix would have to become ad supported.  More recently, I predicted that CTV would become the growth area for TV and a very significant part of networks’ revenue bases.

All of these predictions have come true.  The motivation for these predictions was that I believed that precision targeting of ad impressions would become much more of a driver than achieving reach (the insight and contrary to Byron Sharp and Les Binet thinking).  Who has better data on shopping intentions than Amazon?  CTV is addressable. Netflix knew more about what entertains people than anyone.  All I had to do was push myself to find the predictions that were embedded in those observations.

I encourage all of you to put your insights to the same test.  Ask yourself…

1. If these insights are true, what predictions do they lead to? Then put them on the table for all to inspect.
2. How can you test the implication of the insight to know if the insight is true?
3. If true and based on predicted impact, what different actions should your organization or client undertake to create incremental growth?

Finally, let me suggest that you design the research with the last point in mind…what is the impact that this research can have on incremental growth for the enterprise?  If that is not yet clear, keep refining your research plan.

Your goal? Your research should be shaping the marketing team’s next moves.