INTRODUCTION

Marketing-mix models fundamentally are designed to measure the impact on brand sales of each of the key elements of a brand’s marketing mix. Broadly, marketing-mix modeling consists of statistical analyses, such as multivariate regressions, using sales and marketing time-series data to estimate the impact of various marketing tactics on sales.

Marketers rely on these models to optimize their marketing spending by using them to determine how they should allocate their marketing budget among the various alternative tactics. Once the models are built, it is possible to run alternative scenarios of marketing spending across tactics and optimize that spending as required.

Marketing-mix models came of age in the early 1980s, when supermarket retailers first began using point-of-sales UPC scanners, which provided the granular sales data that the marketing-mix models needed. Store scanner data provided an accurate measurement of sales, because they were based on large sample sizes and could be used as a time-based measure. Causal data (i.e., the marketing drivers of consumer choice) covering the key elements of a typical marketing plan—such as television and print advertising, temporary price reductions, coupons, and in-store merchandising—were readily available for each of the marketing elements used commonly in the 1980s. As a result, marketing-mix models based on aggregated store scanner data quickly became popular.