Analytical and Computational Models for In-Store Shopper Journeys

Retailing is an important part of the supply chain, the link at which money is transferred from the consumer and flows to upstream trading partners. However, retailing has received very little attention from researchers in terms of developing quantitative models that are used in the rest of the supply chain. This paper proposes analytical and computational models based on queuing theory and discrete-event simulation, respectively, for in-store shopper journeys. Specifically, based on empirical data from retail stores, we propose a \(M/G/\infty \) and \({M}_{t}/G/c\) queues in tandem to model the shopping process and the checkout process to characterize key aspects of shopper journeys. Such an analytical model is probably the very first of its kind proposed to mathematically characterize retailing and offers prospects for enabling better understanding and decision-making in this multi-trillion-dollar industry. A key advantage of such analytical models is that they can provide quick answers to questions such as average customer waiting times at checkout for different levels of staffing, which is an important issue in customer service vs staffing cost. A shortcoming of such analytical models is that they are based on assumptions therefore their output has limited accuracy, and these models cannot predict variance in the output. To overcome these shortcomings, this paper presents ShopperSim, which is a stand-alone discrete-event simulation software being developed using SimPy, a Python library, to simulate shopper journeys in a variety of store formats. To ensure practical relevance of these developments, ShopperSim has been programmed to statistically reproduce empirically observed shopping time, basket size, and store area covered – key attributes for shopper journeys. Tests indicate that ShopperSim can reproduce empirically observed statistics of key attributes for shopper journeys. Data generated from ShopperSim indicates that \(M/G/\infty \) and \(M/G/c\) , a much simpler and more tractable tandem queuing model may be adequate for the tested case of a convenience store. In the future, analytical and computational models developed here can be leveraged in several ways for education, training, and operational decision-making in stores, including gamification of these decisions.