Customers in today’s evolving markets are seeking options that best suit their specific needs; consequently, the demand for highly-customized and personalized products has been growing steadily. In order to facilitate customized manufacturing, the structure of the underlying supply chain (SC) needs to be enhanced in terms of flexibility and resilience. In this study, we consider an SC comprising a main manufacturer that produces custom-designed modular-structured products, featured with different design complexity levels. The products have a bill-of-material (BOM) that can be altered in terms of the design of a subset of sub-assemblies and components. It is further assumed that the company collaborates with a group of manufacturers and part suppliers (upstream SC entities), differentiated in terms of capacity, technological capabilities, and cost, along with a group of logistics carriers (downstream entities) distinguished in the sense of their cost and lead time. In other words, in addition to involving customers in the product design, the company also provides different modes of product delivery (e.g., fast, regular, and slow) offered by different logistics carriers. We also incorporate the uncertainty involved in the manufacturing of subassemblies and parts featured with complex designs. More specifically, to reflect the amount of effort required to manufacture complex designs, the production and procurement costs of (highly or moderately complex) items are considered as piece-wise linear functions of their order quantity. Furthermore, the manufacturing capability of upstream entities for producing complex items are assumed uncertain, and modeled as scenarios. More precisely, under some scenarios, the producers/suppliers will not be capable of fulfilling the order of highly/moderately complex items within the promised production lead time due to technological limitations.

To obtain the optimal configuration of the above-mentioned SC, we first develop a deterministic mixed-integer programming (MIP) model that seeks the optimal choice of sub-assembly producers, part suppliers, and logistics carriers in addition to the optimal quantity of procurement, production, and transportation at different echelons. The objective is to minimize the total manufacturing, transportation, and lost-sale cost. Afterward, in order to incorporate the uncertain technological capabilities of producers/suppliers, the above MIP model is reformulated as a two-stage stochastic program (2-SP). In this model, three types of corrective (recourse) actions are considered in order to compensate for the incapability of producers/suppliers in the manufacturing of complex products. These actions correspond to resorting to backup suppliers/producers; purchasing sub-assemblies and components from the open market; and lost-sale, in case none of the first two actions are feasible. Obviously, the cost of all recourse actions mentioned above is substantially higher than the cost of the initial assignment of producers/suppliers that must be set prior to receive full insight in terms of suppliers’ capabilities. We explore a sample average approximation (SAA) scheme to solve the 2-stage stochastic MIP model under different scenario sets. In-depth computational experiments are conducted to validate the proposed models and solution algorithm while showcasing the value of incorporating uncertainty into the SC design problem.