Accounting for Causality When Measuring Sales Lift from Television Advertising

Establishing causality has long been an issue in the advertising-to-sales relationship. The ideal approach in establishing causality is randomized assignment of matched respondents to an exposure and non-exposure group. Given the difficulty in implementing such an experimental approach, observational data has been used in establishing a linkage between advertising and sales. The problem here is that the lack of randomized trials requires controlling for endogeneity (i.e., factors other than advertising related to sales), and accounting for heterogeneity (i.e, variations in consumer characteristics affecting responses to advertising). The authors apply two state-of-the-art methods to single-source data so as to control for endogeneity and account for heterogeneity. Based on this analysis, the authors hope to improve on the use of observational data in accounting for advertising effects in a causal fashion.