Managing spare parts inventory efficiently is essential for automotive aftermarket operations due to unpredictable demand and significant associated costs. Excess inventory leads to increased holding costs, risks of obsolescence, and inefficient use of capital and warehouse space. Conversely, insufficient inventory levels can result in prolonged vehicle downtime, compromised service quality, and decreased customer satisfaction and trust. Effectively managing this balance requires advanced forecasting methods that accurately predict demand variability and reliable inventory control strategies capable of adapting quickly to demand fluctuations, thereby minimizing costs and maximizing efficiency.

This thesis investigates how demand forecasting models combined with prediction intervals (PIs) can improve demand forecasting and inventory control in Volvo Group’s Service Market Logistics (SML), focusing on Volvo Trucks in Sweden. Using real operational data and discrete-event simulation, this study assesses how forecast accuracy and uncertainty affect inventory performance metrics, including service levels and total costs.

The research evaluates how incorporating bootstrap-based prediction intervals at various confidence levels (60%–99%) into Volvo Group’s existing demand forecasts affects inventory management outcomes. Five point-estimation strategies—standard Volvo forecasts, upper-bound forecasts, triangular distribution mean, Beta-PERT distribution mean, and opti standard point were systematically compared using discrete-event simulation. Additionally, the study examines the sensitivity of inventory performance to controlled forecast errors (random, systematic, intermittent) across approximately 14,000 spare parts classified by demand volume, weight, and cost characteristics.

Results indicate that integrating prediction intervals, particularly via Beta-PERT and triangular distributions, substantially reduces inventory costs for low-demand parts—by achieving cost reductions of more than twofold—meaning Volvo’s baseline forecast incurred inventory costs up to 214.7% higher compared to the prediction interval-based approach—while maintaining high service levels. Conversely, conventional forecasting remains preferable for higher-demand parts. This difference arises because high-demand parts exhibit more stable and predictable usage patterns, making conventional point forecasts sufficiently accurate and cost-effective. In contrast, low-demand parts tend to have intermittent and highly variable demand, where prediction intervals better capture uncertainty and help optimize inventory levels by reducing both stockouts and excess inventory.

Systematic overestimation significantly increases inventory costs (+50.7%) with minimal service level improvement, highlighting an asymmetric impact based on forecast error type. The study identified a transition point between 1,000 and 2,000 units annually where prediction intervals stop being beneficial and single-point forecasting becomes more effective.

This research contributes to a practical framework for automotive aftermarket logistics, demonstrating that incorporating forecast uncertainty through prediction intervals significantly improves the efficiency of inventory decision-making, especially in low and variable demand scenarios. These findings have potential applicability beyond automotive contexts, offering insights for industries facing similar inventory management challenges.