Improving the forecasting approach for crude oil sourcing at an oil refining firm: a decision theory perspective

Abstract

Listen

Translate article

Accurate price forecasting is pivotal for making informed sourcing decisions. This study addresses this need by empowering logistics and supply planning managers at a leading European oil refining company through the innovative use of unstructured big data analytics. By applying the CIMO (Context, Intervention, Mechanism, and Outcome) framework, the research refined the oil price forecasting process, which is crucial for optimising sourcing strategies. A novel approach was employed, combining sentiment analysis via deep learning models with traditional autoregressive integrated moving average (ARIMA) models, introducing sentiment as a vital exogenous variable. Through comprehensive stakeholder interviews, the study identified critical gaps and opportunities, leading to the development of a forecasting solution that significantly enhanced predictive accuracy. This advancement eliminated the need for manual interventions by senior decision-makers, streamlining the decision-making process. This research utilises decision theory and also demonstrates the transformative power of unstructured big data analytics in supply chain management. By integrating time series modelling with advanced sentiment analysis, this study provides a blueprint for leveraging technological advancements in rapidly evolving markets. The originality of this work lies in its practical application, showcasing essential capabilities that are increasingly vital in supporting sourcing practices within a volatile global landscape.