A simulation-based hybrid causal predictive framework for stockout risk analysis in supply chain.

Area of Science:

• Operations Research and Management Science
• Supply Chain Analytics
• Causal Inference in Business

Background:

• Stockout risk poses a significant threat to supply chain efficiency and customer satisfaction.
• Existing research often struggles to establish clear causal links due to confounding factors.
• Integrating advanced statistical methods with predictive analytics is crucial for robust analysis.

Purpose of the Study:

• To investigate the causal effect of lead time on stockout risk using a multi-method approach.
• To compare the performance of various causal inference techniques and machine learning models.
• To provide data-driven insights for enhancing operational resilience and inventory management.

Main Methods:

• Employed a framework combining causal inference methods: Propensity Score Matching (PSM), Instrumental Variables (IV-2SLS), Inverse Probability Weighting (IPW), and Doubly Robust Estimation (DRE).
• Integrated machine learning (ML) algorithms (Random Forest, LightGBM) and time series forecasting (Moving Average).
• Utilized a dataset of 20,000 supply chain incidents to estimate Average Treatment Effect (ATE) and evaluate predictive performance.

Main Results:

• Propensity Score Matching (PSM) yielded the most credible ATE (0.882), demonstrating a strong causal relationship between lead time and stockout risk.
• Instrumental Variables analysis showed a weaker, statistically insignificant ATE, indicating potential instrument weakness.
• Random Forest and LightGBM achieved high predictive accuracy (R2=0.25), while Moving Average forecasting effectively captured stockout patterns (R2=0.883).

Conclusions:

• Propensity Score Matching (PSM) is identified as the most robust causal inference technique for this supply chain problem.
• The study highlights the significant causal impact of lead time on stockout risk.
• Integrating causal inference, ML, and time series forecasting offers valuable, actionable insights for proactive inventory management and operational resilience.