Optimization of consumer promotions in the retail sector: a case study

Purpose This study aims to address two key managerial decisions in marketing: allocating the promotional budget and selecting customer groups for each promotion. Design/methodology/approach The authors propose a linear programming model that optimizes these decisions jointly to maximize campaign returns. Rather than assigning promotions to broad, manually defined customer segments, the framework applies machine learning to form data-driven micro-clusters based on customers’ estimated propensities for each promotion type. Findings Computational results show that the mini-batch k-means algorithm can generate up to 20,000 high-quality clusters with limited computational effort, enabling precise and efficient targeting. Moreover, the proposed linear programming model can be solved within reasonable time even at this scale, demonstrating the practical applicability of the solution approach. Practical implications The system aims to improve marketing campaign performance while reducing manual effort and minimizing reliance on user expertise. The system is also designed to accommodate extensive customization. Marketing managers may adjust promotion parameters and specify objectives – such as profit maximization, inventory reduction or customer acquisition – according to current strategic needs. Originality/value Integrating linear optimization with machine learning techniques provides a unified and analytically robust solution to marketing budget allocation, improving promotional effectiveness and resource utilization.