In 1996, while Chief Research Officer at The NPD Group, I co-authored (with Al Baldinger) a paper for the Journal of Advertising Research…”Brand Loyalty, the link between attitudes and behaviors”…that was based on studying the change in brand preferences using the BrandBuilder model for which I developed the math. We studied 27 brands based on 4,071 respondents interviewed twice, one year apart.
This paper generates tremendous readership and academic references, even today. Why is it so relevant today? And, what has changed 20 years later?
So what was the big news?
Research methods (geek alert!). We were the first to prove you could model purchase behaviors towards brands at the individual user level, using constant sum questions*, the statistical Beta Distribution, and modeling attribute ratings against this. This protocol was used in every NPD BrandBuilder study we conducted (we studied over 500 brands). In fact, by measuring each user’s propensities to purchase every significant brand in the category, we were able to accurate predict each brands’ annual penetration, repeat rate, and brand switching patterns (reflective of marketing structure).
We concluded that loyalty, which we defined as a propensity to buy in excess of random purchase rates reinforced by brand beliefs, was more comprehensive than a penetration metric which is merely a by-product of loyalty as we defined it. I had fun debating Andrew Ehrenberg on this.
Brand churn. Loyalty is not to be taken for granted and nearly half of those loyal to your brand in one year (probability of purchase is at least 50%) might forsake it a year later. Worth noting…we could pinpoint WHO was vulnerable and WHO was winnable (called ‘prospects’), by knowing the degree to which brand attitudes were not consistent with current purchase behaviors for each given respondent. In fact, we proved that your ability to conquest a consumer was 5-10 times higher if they were loyal to another brand behaviorally but already had positive thoughts towards your brand. Hence, identification of this “Prospects” segment…who, how big…was critical.
Predictive insights. True loyalty (behavior reinforced by brand attitudes) was a proven basis for predicting directional changes in brand market share, in the aggregate, from one year to the next.
So, which of these insights about brand loyalty and share change is no longer true in 2016? Scroll down…
Trick question! They are all still true today! For example…
Loyalty dynamics. Catalina used frequent shopper data about 6 years ago to prove that while purchase loyalty exists, it DOES shift from one year to the next at similar defection rates to our work. Jannie Hofmeyr from TNS also published evidence that proves churn is significant. Nielsen/Catalina and TRA have both published evidence that people in the middle of the loyalty curve are more persuadable and that targeting them with ads and offers generates significantly more response.
Use of constant sum. Studies I have worked on over the past few years prove that constant sum is still the best way to measure brand preference and probability of purchase.
So what IS different in 2016? ACTIONABILITY AND PREDICTIVE ACCURACY!
Now we can do so much more with these insights! Back in 1996, we lived in a mass marketing world. It is very hard to achieve much lift in targeting efficiencies against a group like persuadables if you need to find a TV show where they cluster. Yes, we proved we could predict who was most likely to defect or to be won over by you, but that was for samples of a few thousand so the insights were strategic but not very actionable tactically.
Today, with DMPs, data driven marketing and real time programmatic ad placement, we can do much more with these insights in a digital, social, mobile age!
Here’s HOW to conduct actionable brand tracking. Use constant sum questioning in your survey, fielded to online panelists whose digital profiling data can be integrated. Model the Prospects and the Vulnerables using lookalike modeling at scale as the survey gives us the truth set and digital profiling variables (including frequent shopper data profiling) in your DMP from those same respondents give us the predictor variable set. We can also see from social media and digital behaviors if brand perceptions and behaviors are changing and build those data streams into our forecasting models.
This might have been science fiction in 1996 but today it is science…merger of insights and data science to be exact.
I’d sum all this up by saying, “Actionable brand tracking…what a concept!”
If you would like to know more about how to create an actionable tracking program in a digital, social, mobile age, or would like a copy of the foundational paper I co-authored, please e-mail me email@example.com.*constant sum questions ask the respondent to allocate 10 points based on their likelihood to buy each brand in their consideration set. Because more points to one brand mean fewer points left over for the others, it is a measure that works better than scale questions for global brand research where different cultures use scales very differently.