Inventory policy design in modern supply chains must balance cost efficiency, service reliability, and responsible resource utilization under significant demand and supply uncertainty. In many real-world supply chains, both customer demand and replenishment lead time exhibit substantial variability, making the design of continuous-review (s, S) inventory policies challenging. Although stochastic inventory models have been widely studied, many existing approaches rely on simplified assumptions or single-objective formulations, which may limit their applicability under simultaneous demand and lead-time uncertainty. This study proposes a data-driven multi-objective simulation–optimization framework for designing sustainable (s, S) inventory policies under dual uncertainty. The framework integrates empirical stochastic modeling, Monte Carlo simulation, and evolutionary multi-objective optimization to evaluate trade-offs between expected inventory cost and service performance. To enhance methodological rigor, statistical reliability control is incorporated into the simulation-based evaluation process to ensure that Pareto dominance relationships are not distorted by simulation noise. Historical operational data are used to estimate probability distributions for demand and lead time, which are incorporated into a stochastic simulation model representing inventory system dynamics. A multi-objective evolutionary algorithm (NSGA-II) is employed to identify Pareto-efficient policy parameters. An empirical case study from a health supplement supply chain demonstrates how the framework identifies efficient replenishment policies under realistic uncertainty conditions. The results reveal structural trade-offs between inventory cost and service level and show that data-driven policy design can improve decision transparency compared with heuristic replenishment rules. The proposed approach provides a structured decision-support tool for selecting replenishment policies that balance service continuity and inventory sustainability in shelf-life-constrained supply chains.
You et al. (Mon,) studied this question.