Demand fulfillment in customer hierarchies with stochastic demand

PDF
Dissertation_Nouri.pdf - Veröffentlichte Version
Download (2MB)

URL:	https://madoc.bib.uni-mannheim.de/61733
URN:	urn:nbn:de:bsz:180-madoc-617330
Dokumenttyp:	Dissertation
Erscheinungsjahr:	2022
Ort der Veröffentlichung:	Mannheim
Hochschule:	Universität Mannheim
Gutachter:	Fleischmann, Moritz
Datum der mündl. Prüfung:	29 September 2021
Sprache der Veröffentlichung:	Englisch
Einrichtung:	Fakultät für Betriebswirtschaftslehre > ABWL u. Logistik (Fleischmann 2009-)
Fachgebiet:	650 Management
Freie Schlagwörter (Englisch):	supply chain management , demand fulfillment , customer hierarchy , make-to-stock production system
Abstract:	Supply scarcity, due to demand or supply fluctuations, is a common issue in make-to-stock production systems. To increase profits when customers are heterogeneous, firms need to decide whether to accept a customer order or reject it in anticipation of more profitable orders, and if accepted, which supplies to use in order to fulfill the order. Such issues are addressed by solving demand fulfillment problems. In order to provide a solution, firms commonly divide their customers into different segments, based on their respective profitability. The available supply is first allocated to the customer segments based on their projected demand information. Then, as customer orders materialize, the allocated quotas are consumed. The customer segments commonly have a multilevel hierarchical structure, which reflects the structure of the sales organization. In this thesis, we study the demand fulfillment problem in make-to-stock production systems, considering such customer hierarchies with stochastic demand. In the hierarchical setting, the available supply is allocated level by level from top to bottom of the hierarchy by multiple planners on different levels. The planners on higher levels of the hierarchy need to make their allocation decisions based on aggregated information, since transmitting all detailed demand information from the bottom to the top of the hierarchy is not generally feasible. In practice, simplistic rules of thumb are applied to deal with this decentralized problem, which lead to sub-optimal results. We aim to provide more effective approaches that result in near-optimal solutions to this decentralized problem. We first consider the single-period problem with a single supply replenishment and focus on identifying critical information for good, decentralized allocation decisions. We propose two decentralized allocation methods, namely a stochastic Theil index approximation and a clustering approach, which provide near-optimal results even for large, complicated hierarchies. Both methods transmit aggregated information about profit heterogeneity and demand uncertainty in the hierarchy, which is missing in the current simplistic rules. Subsequently, we expand our analysis to a multi-period setting, in which periodic supply replenishments are considered and periods are interconnected by inventory or backlog. We consider a periodic setting, meaning that in each period we allow multiple orders from multiple customer segments. We first formalize the centralized problem as a two-stage stochastic dynamic program. Due to the curse of dimensionality, the problem is computationally intractable. Therefore, we propose an approximate dynamic programming heuristic. For the decentralized case, we consider our proposed clustering method and modify it to fit the multi-period setting, relying on the approximate dynamic programming heuristic. Our results show that the proposed heuristics lead to profits very close to the ex-post optimal solution for both centralized and decentralized problems. Finally, we look into the order promising stage and compare different consumption functions, namely partitioned, rule-based nested, and bid price methods. Our results show that nesting leads to performance improvements compared to partitioned consumption. However, for decentralized problems, the improvement resulting from nesting cannot mitigate the profit loss from considerable mis-allocations made by simplistic rules, except for cases with high demand uncertainty or low profit heterogeneity. Moreover, among the nested consumption functions, the bid price approach, which integrates the allocation and consumption stages, leads to a higher performance than the rule-based consumption methods. Altogether, our proposed decentralized methods lead to drastic profit improvements compared to the current simplistic rules for demand fulfillment in customer hierarchies, except for cases with very low shortage or for largely homogeneous customers, where simplistic rules perform similarly well. Applying our advanced methods is especially important when the shortage rate is high or customers are more heterogeneous. Regarding order promising, nesting is more crucial when demand uncertainty is high. The research presented in this thesis was undertaken as part of the project “demand fulfillment in customer hierarchies”. It was funded by the German Research Foundation (DFG) under grant FL738/2-1.