Sourcing and Procurement Cost Allocation in Multi-Division Firms

• 成果类型: Article
• 署名作者: Fang, Fang; Natarajan, Harihara Prasad
• 署名单位: California State University System; California State University Los Angeles; University of Miami
• 刊物名称: PRODUCTION AND OPERATIONS MANAGEMENT
• ISSN/ISSBN: 1059-1478
• DOI: 10.1111/poms.13139
• 发表日期: 2020
• 页码: 767-787
• 关键词: procurement; cost allocation; applied optimization; polyhedral methods

摘要

Through Central Procurement Organizations (CPOs), large firms with multiple divisions have begun adopting a center-led sourcing approach that allows firms to centralize strategic sourcing activities, while permitting decentralized execution by divisions, allowing the firm to leverage large purchase volumes with vendors. This new center-led procurement environment has brought a new decision requirement: How should a CPO select vendors for each division's requirements to minimize the firm's total procurement cost and simultaneously develop a fair and alignment-inducing mechanism to allocate the costs (and savings) of company-wide procurement to the divisions? Past research and current practice have not addressed this linkage between vendor selection and cost allocation in multi-division firms. This work models this sourcing and procurement cost allocation (SPC) problem facing CPOs of large firms as a mixed-integer optimization problem. This model is flexible and can incorporate several commonly-used cost allocation rules. We show that the SPC problem is NP-hard. Therefore, to support practical decision-making in this context, we develop a tailored solution approach for the SPC problem. Our approach enhances the base model by adding strong valid inequalities. We tested our model and its enhancements in an extensive numerical study, solving over 160 instances. Results reveal that (a) the CPO can ensure fair cost allocation at a modest cost (relative to the centralized solution) with our model, and (b) the proposed tailored approach is effective and necessary in solving a wide variety of instances.