Presented by J. Dennis Bender and Ross Link at the American Marketing Association Advanced Research Techniques Forum, June ’96, Beaver Creek, Colorado

Marketing mix models generally include advertising using only a measure of advertising quantity: gross rating points or "GRPs". However, anyone who watches TV knows there is good advertising and bad advertising. Thus, GRPs are not created equal and can not be simply added. Survey research data provides a mechanism to distinguish strong from weak advertising. This presentation discusses the contribution that data has made.

We present a marketing mix model which uses Millward Brown’s awareness index ("AI") as an indication of a particular advertising campaign’s ability to break through the clutter, be noticed and raise branded-advertising awareness. The model is unique in several ways. First, it was one of the first models to integrate behavioral survey data into a scanner data sales response model. Second, it successfully measures and accounts for consumer stockpiling during trade promotions. Finally, it is the only production model we are aware of that directly links advertising awareness to sales results using an empirical Bayesian shrinkage estimator with sign constraints.

In this presentation, we discuss tests done on the model, including holdout sample validation tests. We show the value of the model and the contribution of the survey data to the model’s predictive performance.