Transportation price management for spare parts logistics : a mixed-methods approach using machine learning and qualitative insights

Abstract

Managing transportation costs is crucial for ensuring timely delivery and operational readiness of spare parts in capital-intensive industries. However, volatile shipping rates and fragmented processes can lead to inefficiencies and lost revenue. This thesis presents a mixed-methods transport price management framework that uses machine learning and qualitative stakeholder insights to improve cost transparency and decision support. A Random Forest model is trained on 71,554 courier and 2,388 ocean freight shipments from Finnish and German hubs, including actual and volumetric weight, container configuration, provider zones, and great-circle distances. The model explains 96% of the variance in courier costs and 83% in ocean freight, indicating strong predictive performance. Semi-structured interviews with relevant stakeholders revealed key organisational challenges, including inconsistent billing practices, manual dimensioning workflows, outdated cost calculators, and unclear policy ownership. The thesis recommends implementing a centralised goods policy with dedicated oversight, automated cost-calculation tools linked to invoicing systems, and targeted training modules to standardise practices. By incorporating predictive analytics into a structured governance model, the proposed framework enables more accurate budgeting, optimised mode selection, and improved service levels in spare parts logistics. The thesis demonstrates how non-linear machine learning techniques can replace traditional cost models and provides a scalable roadmap for data-driven transportation price management.