The energy sector faces persistent challenges in managing spare parts inventory due to demand volatility, operational complexity, and capital intensity. Our Sponsor, one of the world’s largest energy companies, sought to optimize inventory by aligning stock levels more closely with business demand, with a goal of reducing excess holdings and improving service reliability. To address this challenge, we conducted a comprehensive analysis of over 40,000 materials across nine business units, segmenting them into four categories based on demand intermittence and variability metrics. We then tested seven machine learning models across the categories identified, ultimately selecting the two best-performing ones. For materials lacking sufficient historical data required for machine learning, we implemented a pragmatic alternative using average consumption. This two-pronged approach revealed a nearly 27% inventory reduction opportunity while achieving a 95% target service level. Our findings demonstrate that a dynamic, demand driven inventory policy can unlock substantial working capital, improve asset readiness, and enhance operational resilience, providing a scalable framework for enterprise-wide inventory optimization.
