Uncertainty-Aware Thinking-Model Planning for Dynamic Supply Chain Network Design

Strategic facility activation and operational inventory allocation are coupled sequential decisions under demand volatility. This study develops an uncertainty-aware thinking-model decision evaluator (UA-TD) for multi-period supply chain network design. A bootstrapped one-step demand predictor and empirical multivariate innovations generate scenario paths; exact cost accounting and a feasible shipment recourse rule evaluate enumerated facility sequences. An expectation–conditional-value-at-risk objective extends the candidate-sequence evaluation of the Think4SCND thinking model, retaining the first action and replanning after each observed demand. Fully synthetic, explicitly specified experiments include three warehouses, four customer zones, ten periods, and 180 held-out episodes per demand regime. Relative to risk-neutral ensemble planning, UA-TD increases mean fill rate by 0.77, 0.85, and 0.52 percentage points in normal, volatile, and disruption conditions, respectively, with paired bootstrap intervals excluding zero. Mean episode cost is 0.68% higher in normal conditions and 0.29% and 0.20% lower in volatile and disrupted conditions. The evidence supports a measurable service–cost trade-off, not uniform dominance. The study contributes an inspectable, risk-sensitive planning module and a fully reproducible evaluation protocol for dynamic facility and inventory decisions.