A stochastic programming model with endogenous uncertainty for proactive supplier risk mitigation of low-volume-high-value manufacturers considering decision-dependent supplier performance

Article Status: Published
Publication Year: 2022
Rui Zhou, Bhuiyan, Tanveer Hossain, Michael Sherwin, and Medal, H. (2022).. Zhou, Rui, Tanveer Hossain Bhuiyan, Hugh R. Medal, Michael D. Sherwin, and Dong Yang. "A stochastic programming model with endogenous uncertainty for selecting supplier development programs to proactively mitigate supplier risk." Omega 107 (2022): 102542.
[External Link]

Poor supplier performance can result in delays that disrupt manufacturing operations. By proactively managing supplier performance, the likelihood and severity of supplier risk can be minimized. In this paper, we study the problem of selecting optimal supplier development programs (SDPs) to improve suppliers’ performance with a limited budget to proactively reduce supplier risks for a manufacturer. A key feature of our research is that it incorporates the uncertainty in supplier performance in response to SDPs selection decisions. This uncertainty is endogenous (decision-dependent), as the probability of supplier performance depends on the selection of SDPs, which introduces modeling and algorithmic challenges. We formulate this problem as a two-stage stochastic program with decision-dependent uncertainty. We implement a sample-based greedy algorithm and an accelerated Benders’ decomposition method to solve the developed model. We evaluate our methodology using the numerical cases of four low-volume, high-value manufacturing firms. The results provide insights into the effects of the budget amount and of the number of SDPs on the firm’s expected profit. Numerical experiments demonstrate that an increase in budget results in profit growth, e.g., 5.09% profit growth for one firm. At a lower budget level, increasing the number of available SDPs results in more profit growth. The results also demonstrate the significance of considering uncertainty in supplier performance and considering multiple supplier risks for the firm. In addition, computational experiments demonstrate that our algorithms, especially our greedy approximation algorithm, can solve large-sized problems in a reasonable time.