Incorporating risk aversion and time preference into omnichannel retail operations considering assortment and inventory optimization

An omnichannel retail network includes an omnichannel retailer (o-retailer) determining which products among the candidate set are offered via an online or offline channel, and then procuring these products from the supplier and distributing them from a distribution center (DC) to physical stores for serving customers. The o-retailer adopts either the distribution center or physical stores to fulfill online orders. This study considers an integrated approach to manage assortment planning, inventory control and e-fulfillment problems and develops a distributionally robust optimization model, in which the distribution of uncertain demand is only partially available in advance. In the proposed model, the worst-case mean-Conditional Value-at-Risk (WMCVaR) is formulated as the objective function that makes a trade-off between the expected profit and the risk, and a quasi-hyperbolic discounting function is adopted to denote the customers' time preference. Furthermore, to overcome model solvability obstacle caused by imprecise probability distributions, the box ambiguity set is applied to derive computationally tractable counterparts. Numerical studies are conducted to investigate the validity and efficiency of our proposed model. Some useful managerial insights and implications for o-retailers are generated through the analyses of computational results.