A Bayesian Dynamic Linear Model for Brand Equity
Co authors: Elea Feit and Eric Bradlow
This paper presents a statistical model to measure brand equity as it changes over time, and gauges the impact of changing brand equity on consumers’ product choices. Our model extends traditional models of brand equity by using weekly UPC-level product choices (nested within brands in a hierarchical way) to estimate a brand’s dynamic equity, while also (in a new to the literature way) accounting for dynamic brand equity’s affect on marketing mix coefficients. Due to the potential for both marketing mix and price endogeneity, we introduce a two-level copula structure as price and marketing enter our Bayesian models at different levels. We apply the proposed model in an empirical application using a panel data set of households’ purchases in the diet soda category, demonstrating that brand equity has a measurable effect on price sensitivity and other marketing mix coefficients. We utilize our model to optimize the timing and depth of marketing and to derive optimal weekly prices comparing those results to simpler benchmark models. As we show, this extended model provides firms and researchers with a more comprehensive view of brand equity, and leads to novel strategies for building and exploiting the strength of a brand.