风险规避型全渠道零售商多周期产品组合与库存联合优化

Multi-period assortment and inventory stochastic optimization for risk-averse omnichannel retailers

基于多项式Logit(multinomial logit,MNL)选择模型,本文研究风险规避型全渠道零售商多周期产品组合与库存联合优化问题,建立了具有服务水平约束的多周期随机优化模型并提出相应的求解方法,通过数值算例验证了所构建模型和求解方法的有效性。研究结果表明,基于均值-CVaR的随机优化模型,能够使全渠道零售商更好地权衡利润与风险,并且使利润表现出更好的鲁棒性,进而满足零售商的风险规避需求。针对全渠道零售商实体店容量约束以及履单中心容量约束,零售商的相关决策会随着约束的改变而变化。两种渠道的存在在保证产品多样性方面具有一定优势。通过探究消费者的选择行为发现,当消费者群的考虑集完全不重叠时产品被选率是最高的。采用混合订单履行策略优于单一履行策略,且有助于实现产品的多样化,保证产品组合决策的稳定性与灵活性。

As technologies evolve and consumers′preferences change,more and more retailers are adopting omnichannel retailing systems to attract consumers and improve their profitability.Omnichannel retailing unlocks numerous opportunities for a retailer to leverage its brick-and-mortar store network to support its online sales.However,retailers are still confronted with challenges and constraints.An omnichannel retailer faces a set of complex decisions that affects its profitability and competitiveness.The assortment decision is one of the most important decisions.The assortment decision will determine which items will offer to consumers through brick-and-mortar stores,online channels or via both channels.Research shows that a larger assortment of items offered in retailers′online channels with no shelf space restrictions will help to increase sales.Furthermore,the products offered in the online channel will affect the demand of both online and offline channels.Besides the assortment decision,retailers have to determine how much inventory to stock for each product.The limited space will impact the inventory decisions.Meanwhile,retailers are facing uncertain demands.Hence,how to make joint decisions on assortment and inventory given this situation so as to minimize the retailer′s profit uncertainty is essential for risk-averse omnichannel retailers.In addition,the omnichannel retailers require to develop new cost-efficient logistics business models to integrate different channels seamlessly when moving to omnichannel operations.Therefore,choosing appropriate order fulfillment strategies and deploying an efficient omnichannel distribution network have become important issues for omnichannel retailers that require urgent solutions.Conditional value-at-risk(CVaR)is used to measure the retailer′s profit uncertainty.The mean-CVaR is applied to make trade-offs between profit and risk.Because the set of products offered influences each product′s demand via both online and offline channels,the multinomial logit(MNL)model is adopted for this research to describe the selection processes of consumers in a realistic manner.A multi-period stochastic optimization model for determining assortment and inventory is proposed under uncertain demand and capacity limits with a service level constraint.The objective of the proposed model is to maximize the retailer′s profit while decreasing the risk of uncertainty.The key events in the multi-period can be characterized as occurring before the selling season,during the selling season,and after the selling season.Before the selling season,the omnichannel retailer chooses offer sets for brick-and-mortar stores and the online channel,and orders products from the supplier.During the selling season,the distribution center(DC)will replenish the inventory in brick-and-mortar stores and the fulfillment center.And the online orders are fulfilled via an“integrated fulfillment”strategy.After the selling season,the residual inventory is disposed of and a salvage value is obtained.Since considering stochastic demand,the Monte Carlo method is used to simulate different demand scenarios.In this study,the MNL model and the chance constraints result in a nonlinear model.In order to solve the nonlinear model,this paper applies linearization techniques to transform the nonlinear model into an equivalent solvable mixed integer linear programming model.Through numerical studies,the applicability of the proposed model is investigated and some important managerial insights are obtained.The results demonstrate that the mean-CVaR model efficiently prevents fluctuations in the expected profit in the supply chain and hedges against uncertainties to avoid high-risk solutions.Although the expected profit under the mean-CVaR model is slightly lower than that of the risk-neutral model,the standard deviation of the profit is significantly reduced.Therefore,the mean-CVaR model has the ability to make trade-offs between optimality and robustness.In this study,the sensitivity analyses show that the product variety increases when capacities increase in the stores and the fulfillment center,and it grows more quickly as the capacity level of brick-and-mortar stores increasing.The expected profit decreases as the demand variance increases.Although both the variety of products stocked in online and offline channels decrease as demand uncertainty increases,the total variety of products offered in both channels does not decrease.This finding implies that the existence of two channels can ensure that the variety of products does not decrease.By exploring consumer choice behavior,it is found that the product selection rate is the highest when the consumers′consideration sets are completely non-overlapping.Concurrently,a wider assortment increases the cannibalization effect.Regardless of the changes in the unit order fulfillment cost,the integrated fulfillment strategy is better than a single order fulfillment.Under the integrated fulfillment strategy,the assortment decision is not affected by the unit order fulfillment cost and can provide a greater variety of products;further,the assortment decision exhibits a certain degree of stability and flexibility.