Forecasting spare parts demand: a case study at an indonesian heavy equipment company

Abstract

Forecasting spare parts demand is a common issue dealt by inventory managers at maintenance service organization. The large number of items held in stocks and the random demand occurrences make most of the difficulties. An indonesian heavy equipment company targets to advance its forecast accuracy. Accordingly, this study has two main goals. Firstly, all stock keeping units (skus) are classified based on their demand patterns, utilizing their average inter-demand interval (adi) and squared coefficient of variation (cv2) of demand sizes as the classifiers. After that, four simple forecasting methods are applied to each demand class and the best forecasting method in term of its forecast errors is chosen. Evaluation of forecast accuracy is made by means of the mean absolute scaled error (mase), mad-to-mean ratio, and percentage best (pbt). The forecasting competition results show the dominance of syntetos-boylan approximation for erratic, smooth, and intermittent demand, and simple moving average for lumpy demand