Robust Optimization for Aggregate Transportation-Mode Allocation in Pharmaceutical Supply Chains under Delivery-Time Uncertainty: Evidence from SCMS Shipment Data
Year: 2026
Type: article
Abstract: Pharmaceutical transportation-mode selection requires balancing freight cost and delivery-delay risk. This study develops an empirically calibrated robust optimization framework for aggregate pharmaceutical shipment allocation across modes, considering cost, delay risk, and a minimum quota of high-value shipments on reliable modes. The model uses 6,804 USAID Supply Chain Management System (SCMS) shipment records, mainly HIV-related commodities delivered during 2006–2015. Mode-specific weight-dependent freight-cost functions are validated against recorded expenditure; delay uncertainty is represented by a budgeted set calibrated from bootstrap estimation errors of mode-level mean positive delays. Ocean transport has the most severe upper tail (90th-percentile delay: 31 days), while ground transport has the highest lateness rate (20.7%). At Γ = 4, ocean transport is removed from the allocation; relative to Γ = 0, estimated freight cost increases by 1.9%, aggregate expected positive delay decreases by 11.7%, and the estimated on-time delivery rate rises from 86.3% to 86.8%, corresponding to approximately 33 additional consignments classified as on time among 6,146 shipments. Across 3,000 Monte Carlo scenarios and tested parameter perturbations, the robust allocation yields no greater delay than the deterministic allocation. In the 2013–2015 holdout, the allocations coincide; across four rolling-origin evaluations, average expected delay is 0.7% higher for the robust allocation. With composition-adjusted delay estimates, the ocean share remains 14.6% across Γ = 0–4, compared with 0% under the baseline robust allocation at Γ = 4. These results show that the cost–delay trade-off depends on the stability of modal delay patterns and shipment composition, highlighting the value of periodic recalibration. Findings are specific to one donor program.
Source: Figshare
Author: Javad Behnamian
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7236326838
