Data-Driven Spare Parts Classification and Decision Support : A Case Study at Volvo Trucks

Sammanfattning

Maintenance spare parts are essential for production uptime in asset-intensive manufacturing, yet managing them remains a practical challenge due to intermittent demand and data scattered across organisational functions. Although multi-criteria classification methods such as ABC, XYZ and VEDhave been proposed in the literature, most studies focus on the classification logic itself and give limited attention to data integration and operational decision support. This thesis addresses that gap through a case study at Volvo Trucks in Tuve, Gothenburg, where data from multiple sources were consolidated into a dimensional model and an integrated ABCXYZ-VED classification was applied to 19,142 unique spare part numbers. The VED dimension wasoperationalised through an AHP-based scoring model, and the results were made available through a Power BI prototype that provides a consolidated view of the inventory and supports counting prioritisation, dead stock review and storage location analysis. Three findings stand out. First, 1.79% of items account for nearly 80% of consumption value, while 95.85% of items fall into class C, leaving ABC analysis with limited di!erentiating power for the bulk of the portfolio. Second, 75.52% of all registered part numbers have no recorded demand during the observation window, which means that criticality rather than consumption history must guide inventory decisions for the majority of items. Third, 62.8% of items are linked to machines that lack a criticality classification, revealing an organisational gap in master data that directly limits the reach of the scoring model. Despite these constraints, 12,345 stocked items were assigned to priority groups, where the majority of the highest-priority items are low-value parts that would be invisible under value-based classification alone. In practical terms, this gives warehouse personnel a basis for decisions that previously depended largely on individual judgement. The study shows that bridging fragmented data sources is necessary but not su"cient for multi-criteria classification. The underlying master data, particularly machine-level criticality assessments, must be maintained as a continuous organisational practice.