Using Bayes Nets for Market Share Driver Analyses

Presented at the 4th Annual BayesiaLab Conference in Nashville on September 29, 2016.

Abstract

One of the more common strategic requests directed to business analytic teams is to analyze market research or related data to better understand the drivers of customer loyalty, satisfaction, or in this case, share of wallet. Using data from an extensive survey in the fresh-squeezed orange juice category, we compare and contrast direct-effect-only models (in this case, multiple regression and Dirichlet regression) with an Augmented Naïve Bayes model. The Augmented Naïve Bayes model, with its captured indirect effects, results in meaningful differences for these kinds of analyses and can change the strategic counsel and recommendations delivered to senior leadership, leading to more effective business strategies and programs.

Presentation Video

About the Presenters

Bob Wood is Senior Director of Consumer Analytics at Merkle, supporting global brands through consumer insights and advanced analytics. He has an MS in Applied Math from BYU, MS in Statistics from Wichita State, an MBA from the Marriott School of Management, and currently a doctoral candidate in consumer psychology at Wichita State.

Dr. Toshi Yumoto is Director—Advanced Methods & Research at Merkle with over 15 years of experience in developing quantitative approaches that help organizations drive customer-centric strategies. Dr. Yumoto has been published in over 50 journals and received his Ph.D. from the University of Maryland College Park in Measurement, Statistics, and Evaluation.