Using transaction data and product margins to optimise weekly flyers

Abstract

Listen

Translate article

Customers increasingly expect companies to understand their wants and needs and to market to those desires. Unfortunately, such levels of personalisation can be difficult to accomplish for traditional brick-and-mortar retailers, particularly when the fixed cost of personalised marketing is significant. This paper considers methods for providing customised promotions to customers in the form of weekly flyers. We consider how multiple versions of a weekly flyer can be used by a retailer, which products should be included in each flyer version, and how customer preferences, markup, and inventory considerations impact these decisions. Specifically, transaction histories are used to estimate customer preferences for various product offerings using market basket analysis. These probabilities are then incorporated into an optimisation model for grouping customers into market segments and presenting option sets within each segment that maximise expected marginal profits across the flyers. A heuristic is proposed for solving larger problems and evaluated against lower and upper bounds on the optimal profit. Computational results indicate that leveraging customer transaction data to optimise the product selection and assignment of four unique flyers can increase profit by 7.7% over the optimal single-flyer solution.