Collaborative Research: Integrating Complex Choice Behavior into Assortment, Inventory, and Pricing Decisions

提供多样化的产品分类给公司带来了许多好处，比如更大的市场份额，更高的客户满意度和更多的重复购买。然而，确定正确的分类品种是困难的，因为分类中的产品可能是互补的或替代的。因此，每种产品的需求取决于该分类中所有其他产品的可用性，使库存计划和定价决策复杂化。多样化的产品分类也意味着需求分散在更多的产品中，这可能导致安全库存增加和其他额外的运营费用。考虑到这些复杂的权衡，公司应该如何处理产品多样性和客户选择行为的运营后果？该项目的目标是开发模型和算法，使决策者能够提供正确的产品分类，库存适量的库存，收取正确的价格并向正确的客户销售，同时在做出这些决定时考虑到客户的选择行为。通过研究最先进的选择模型来捕捉客户的选择行为，并将这些选择模型纳入关键的运营管理问题，该项目提供了一个综合框架来解决分类计划、库存管理、定价和产品个性化决策。该项目涵盖了适用于众多行业的各种问题，重点是分类计划、库存管理、定价和产品个性化决策。(a)项目的一个重点是在库存不是限制问题的情况下作出分类决定，这在销售不需要消耗实物库存的产品时是适当的。该项目将开发精确和近似的算法，以找到在各种选择模型下提供的正确产品分类。(b)提供的分类决定了对不同产品的需求，表明分类决策与库存决策相互作用。该项目将研究联合分类报价和库存决策的模型，以及固定库存可用性下分类报价决策的模型。这些模型的标准公式需要为每种可能的分类提供一个单独的决策变量，这可能会有太多。目标是在不同的选择模型下开发紧凑的配方，并研究最优分类的结构特性。(c)该项目涉及顾客在提供的产品中进行选择时的定价决定。本项目将在多种选择模型下研究这类定价问题，并纳入对价格的各种约束。(d)客户实时数据的可得性的发展使向每个客户提供个性化的分类产品成为可能。挑战在于快速评估每个客户的选择行为，并根据选择行为和库存可用性实时计算出最优的分类。该项目将研究具有性能保证的分类个性化模型，并研究如何纳入客户预测。(e)在客户选择下建立模型的一个重要步骤是对驱动选择行为的选择模型的估计。该项目将研究如何从销售数据中估计选择模型。当数据有限时，在使用复杂模型提高预测精度和对有限数据过度拟合复杂选择模型之间存在权衡。一个重要的目标是推导出健壮的标准来平衡这些权衡。

Offering a diverse product assortment brings numerous benefits to a firm, such as larger market share, increased customer satisfaction and more repeat purchases. However, determining the right assortment variety is difficult because products within the assortment maybe complementary or substitutes. Thus, the demand of each product depends on the availability of all other products in the assortment, complicating inventory planning and pricing decisions. A diverse product assortment also means that the demand is dispersed among a larger number of products, which may result in increased safety stocks and other additional operational expenses. Given these complex tradeoffs, how should a firm deal with the operational consequences of product variety and customer choice behavior? The goal of this project is to develop models and algorithms that allow decision-makers to offer the right assortment of products, stock the right amount of inventory, charge the right prices and sell to the right customer, while taking the customer choice behavior into consideration as they make these decisions. By studying the state-of-the-art choice models to capture customer choice behavior and by incorporating these choice models into critical operations management problems, the project offers an integrative framework to address assortment planning, inventory management, pricing and product personalization decisions.The project covers various problems that apply to numerous industries, with focus on assortment planning, inventory management, pricing, and product personalization decisions. (a) One focus of the project is assortment decisions when inventories are not a limiting concern, which is appropriate when selling products that do not require consumption of physical inventories. The project will develop exact and approximate algorithms to find the right assortment of products to offer under various choice models. (b) Offered assortment determines the demands for the different products, indicating that assortment decisions interact with stocking decisions. The project will study models that make joint assortment offer and stocking decisions, as well as models that make assortment offer decisions under fixed inventory availability. Standard formulations of these models require a separate decision variable for each possible assortment, which can get too many. The goal is to develop compact formulations under different choice models and investigate the structural properties of the optimal assortment. (c) The project covers pricing decisions when customers choose among the offered products. The project will investigate such pricing problems under a variety of choice models and incorporate various constraints on the prices. (d) Developments in availability of real-time data on customers allow personalizing the assortment offering to each customer. The challenge is to quickly assess the choice behavior of each customer and compute the optimal assortment to offer in real time based on choice behavior and inventory availability. The project will study assortment personalization models with performance guarantees and investigate how to incorporate customer forecasts. (e) An important step in building models under customer choice is the estimation of the choice model that drives the choice behavior. The project will investigate how to estimate choice models from sales data. When data is limited, there is a tradeoff between increasing the predictive accuracy by using complex models and over-fitting complex choice models to limited data. An important goal is to derive robust criteria to balance the tradeoffs.