Predictive quality assurance for logistics provider selection: an AI–MCDM framework

Abstract

Quality assurance has evolved from being a competitive advantage to a baseline requirement in logistics services. Traditional multi-criteria decision-making (MCDM) approaches to selecting logistics service providers (LSPs) rely heavily on historical data and expert judgment, which may be unreliable in today’s volatile markets. This paper introduces an AI-enhanced MCDM framework that transforms predictive distributions of key quality indicators—such as on-time delivery, lead-time variability, damage rate, CO₂ emissions, and cost—into risk-adjusted preference scores. These scores are then aggregated using established methods (TOPSIS, PROMETHEE, VIKOR) to generate forward-looking and risk-aware rankings. A PRISMA-guided systematic literature review (2019–2025) highlights three major gaps in the field: limited integration of predictive analytics into MCDM, insufficient treatment of tail risks and data drift, and weak attention to explainability. The proposed framework addresses these shortcomings and is illustrated through a case study that demonstrates how predictive, risk-sensitive rankings can enhance robustness and decision quality in LSP selection.