Latent segmentation using store-level scanner data

Purpose - The purpose of this paper is to incorporate explicitly consumer heterogeneity into market response models estimated with store-level scanner-data. Design/methodology/approach - Latent structures in market response to a product category using aggregated scanner data registered by a supermarket are identified. Specifically, latent consumer segments with diverse preferences towards brands and different responses to marketing stimuli from data consisting of daily marketing actions (i.e. price, promotions, advertising, etc.) and sales of competing brands are identified. Findings - The existence of different latent segments with diverse preferences and response patterns to marketing stimuli were detected. More specifically, the fit of the statistical analysis for the different model possibilities made it possible to identify four market segments. It was also found that the intrinsic brand attractiveness as a measure of consumer brand preference is different between segments. Finally, the price sensitivity is also different between segments. Research limitations/implications - The time cost necessary to obtain the parameter estimates is too high, which is usual in the models estimated with iterative EM algorithms. Practical implications - This work deepens one's knowledge of the identification and selection of latent market structures, specifically latent segments with different purchase patterns and behaviours. The possibility of developing the analysis with aggregated data at the store level increases the potential utility for academics and marketing managers. Originality/value - Although most applications use weekly data, this proposal models daily fluctuations in sales - as a result, making it possible to obtain consumer segments based on daily changes.