Forecasting demand during and after supply chain disruptions using a shock smoother ETS

Abstract

Recent events such as Brexit, the Russian invasion of Ukraine, and the COVID-19 pandemic have highlighted the challenges supply chains face in accurately forecasting demand during and after major disruptions. Traditional methods, which typically perform well under normal conditions, often struggle to provide reliable forecasts when demand is disrupted. As a result, many decision makers rely on their subjective judgment rather than statistical models for demand planning. However, accurate forecasting remains crucial, especially in times of disruption. To address this issue, we make two key contributions. First, using both simulated and real world data, we evaluate several traditional forecasting methods, assessing their overall performance and effectiveness across different phases of disruption. This analysis highlights their limitations in handling disrupted demand patterns. Second, we propose a shock-smoothing model, which is a modification of the single source of error state-space model underlying exponential smoothing (ETS) to include additional components that account for the disruption periods. Our findings demonstrate that the proposed model improves overall forecasting accuracy and maintains greater resilience across individual phases of disruption, positioning it as a potential valuable tool for enabling data-driven demand planning both during and after disruptions.