Analyzing Procurement Data for Cost Saving Application

Abstract

In an increasingly data-driven business environment, procurement analytics plays a critical role in optimizing costs and improving supply chain efficiency. This thesis examines the development and implementation of the Lifecycle Cost Management (LCM) tool at Caterpillar Inc., a global leader in heavy equipment manufacturing. Given Caterpillar's decentralized procurement structure, managing cost-saving initiatives across its 150 facilities (Caterpillar | Caterpillar Frequently Asked Questions (FAQs), n.d.) and 28,000 suppliers (Caterpillar | Caterpillar at a Glance, n.d.) poses a significant challenge. The LCM tool leverages machine learning models to identify overpriced purchase orders (POs) and generate actionable cost-saving opportunities. This study explores the methodology used to enhance LCM's predictive capabilities, including data sourcing and cleaning, feature engineering, model selection, and validation. Various regression models, clustering techniques, and machine learning algorithms, such as Random Forest and XGBoost, are tested to identify cost outliers. A validation process is implemented to ensure that flagged outliers are cost-saving opportunities appropriate for execution. Beyond technical development, the thesis addresses the processes of digital tool adoption within Caterpillar’s procurement teams. A change management approach is employed, incorporating buyer interviews, stakeholder engagement, and iterative user experience (UX) improvements. Through case studies, the study highlights the machine learning model performance and tangible financial impact of LCM. The LCM tool has identified more than $100M data-driven potential savings, and hopes to realize 20% of the savings. Because Caterpillar’s procurement contracts are often long-term, these savings can be considered perpetual. Findings indicate that while machine learning models effectively identify cost outliers, their success is contingent on robust data governance, stakeholder buy-in, and integration into procurement workflows. The study underscores the importance of data management, organizational alignment, and continuous refinement of digital procurement tools. Future work recommendations are enhancing data infrastructure, integrating AI-driven contract management and analysis, and refining cost estimation methodologies. The insights gained contribute to the broader application of procurement analytics and digital transformation in manufacturing enterprises.