Assessing the expected impact of logistics risks on supply chain performance

Abstract

This study develops an integrated framework that predicts logistics risks and quantifies their expected impact on supply chain performance. Existing research largely focuses on estimating risk occurrence but rarely evaluates both the likelihood and the potential consequences of disruptions. To address this gap, the study links machine learning (ML)-based risk prediction with an Expected Risk Impact (ERI) assessment to support more informed and impact-oriented decision-making. The proposed framework combines Random Forest-based prediction of three logistics risks (customs risk, shipment delay risk and route disruption risk) with an ERI calculation derived by integrating predicted probabilities and independent severity measures. The analysis uses a real-world dataset of shipment operations in Southern California from 2021 to 2024. ERI distributions are examined to identify high-impact shipments, and coefficient-based interpretability techniques are applied to determine key risk drivers. The results indicate that the models provide reliable risk probability estimates, while the ERI assessment reveals a highly skewed distribution of impacts, with a small subset of shipments accounting for disproportionately high expected risk exposure. Port congestion, handling operations, supplier reliability and route-related factors emerge as dominant contributors to high-impact risks. The findings enable logistics managers to prioritize high-impact shipments, allocate resources more efficiently and design targeted mitigation strategies, thereby reducing operational disruptions and improving supply chain stability. This study advances logistics risk management by bridging ML-based risk prediction with impact-oriented assessment. The integrated framework moves beyond occurrence-focused approaches and offers a more comprehensive basis for resilient supply chain decision support.