Demand forecasting of spare parts – case study from automotive industry

Abstract

The purpose of this thesis is to find suitable methods in forecasting the intermittent demand for spare parts and in an automotive company. Four existing methods, from simple to exclusively developed for forecasting intermittent demand items, were tested to determine the best-performing method. The optimal method was chosen on the basis of which model provides the lowest level of error against the actual demand. Demand pattern was classified into 4 types of demand pattern, namely erratic, lumpy, smooth and intermittent, by the degree of intermittence and degree of erraticness. Dataset extracted from sales data of 5 operating markets in the period 2013-2018, on monthly aggregation level. Demand was forecasted using forecasting methods such as Moving Average, Exponential Smoothing, Croston’s method, and Syntetos-Boylan Approximation. The study’s result proved that the simple Moving Average is not a good approach in forecasting items with intermittent demand. On the other hand, Syntetos-Boylan Approximation and Exponential are the 2 methods that provided the best accuracy, depending on demand patterns and markets. The findings also proposed a forecasting framework to the case of the company.