MIT SCM Symposium 2024

In response to the challenges of forecasting demand for drilling bits, this capstone project aimed to enhance the accuracy and efficiency of demand prediction for a global technology company in the energy sector. The goal was to replace outdated manual forecasting methods with automated causal and time-series models, optimizing the company’s Sales and Operations Planning (S&OP) processes in preparation for larger automation within the company’s Integrated Business Planning software (IBP). Utilizing historical data on rig counts and market shares, these models predicted future bit runs and estimated associated revenues with significantly improved accuracy, reducing global Mean Absolute Percentage Error (MAPE) from 6–8% to 2–3%. The analysis revealed that the performance of these models varied considerably among different geographic units (“geounits”), highlighting the necessity for customized forecasting strategies. Importantly, the best-performing models correlated with geounits’ business models, whether rental or bulk sales. Simpler time-series models frequently outperformed more complex causal models, suggesting that complexity does not always yield better forecasting results. This project streamlined the forecasting process and laid a strong foundation for ongoing improvement and adaptation to emerging market conditions. The integration of these models into the company’s S&OP practices represents a significant step forward in leveraging technological advancements to maintain a competitive edge in the energy sector.