Managing demand volatility during unplanned events with sentiment analysis: a case study of the COVID-19 pandemic

Abstract

Unplanned events such as natural disasters or epidemic outbreaks are usually accompanied by supply chain disruption and highly volatile markets. Besides, the recent COVID-19 crisis has shown that existing artificial intelligence systems and data analytics models, which normally provide valuable support in demand forecasting, have not been able to manage demand volatility. This study contributes addressing this issue and aims to determine whether sentiments conveyed by news media influence consumer behavior. It provides a case study conducted in three steps: (1) data were collected and prepared; (2) a sentiment analysis model was developed; and (3) a statistical analysis was performed to analyze the correlation between sentiments in news and drug consumption during the COVID-19 crisis. Findings highlighted a strong positive correlation between sentiments in news and consumption variability. They therefore suggest that sentiments in news have strong predictive power for demand forecasting in unplanned situations.