A mixed methods approach to analyze and predict supply disruptions by combining causal inference and deep learning

In today's complex supply networks, disruptions caused by diverse events represent major unknown operational conditions and risk factors requiring research on supply chain risk management. Hereby, operations management (OM) researchers are traditionally focusing on defining analytical and mathematical risk mitigation strategies characterized by a variety of different constraints, such as capacities, timeliness, or relations, leading to limited transferability into practice. Given the complexity of modern supply chains we argue that a data-driven approach to supply chain risk management enhances the assessment and comparability of risk mitigation strategies. Thereby an approach that allows conclusions about causal relationships between supply chain interventions and potential outcomes is of particular importance. Inspired by deep learning as well as causal learning theory and following a design science research approach, we design an analytical model that can predict supply disruptions based on external and internal data and quantify causal effects on delivery reliability. The model is evaluated in a single case study using data from a large first tier automotive supplier. The results show that the model has a high predictive performance and can learn causalities from observed data and analyze interventions effecting supply disruptions. The identification of causal relationships offers the potential to identify lacking supplier relationships and consequently bundle supply chain risk management activities. Building on the empirical and analytical insights, we discuss implications for both theory and practice.