Developing a Spare Parts Demand Forecasting System

Abstract

The focus of this study is on developing a decision support system (DSS) in order to forecast spare parts demand for a company producing high technology products in Turkey. The company is one of the world’s leading original design manufacturers in the field of consumer electronics and white goods. Accurate forecasts of customer demand for preliminary products and spare parts play an important role in order to reduce costs and increase customer satisfaction. Currently, the company’s forecasting system is based on personnel experience and a statistical approach, which lacks the ability of capturing demand data behaviour. The approach followed results in an increased forecasting error, thus increases production costs, results in lack of spare parts, and decreases customer satisfaction. The aim of this project is to develop a DSS to minimize the forecasting error; therefore help the company develop a policy for optimizing the stock levels kept, reducing costs and increasing customer satisfaction. In order to understand the behaviour of customer demand of spare parts, the company’s television products are chosen for the pilot study, since these products are highly influenced by rapid technological changes and changes in the product models. The spare parts are classified into different groups, using ABC analysis, in order to develop a forecasting model for each group. In the solution methodology part, three different statistical methodologies for the forecasting process were respectively studied; Winter’s, Double Exponential Smoothing and Moving Average Methods. Winter’s Method is used for the data which exhibit trend and seasonality, Double Exponential Smoothing is used for the data which exhibit trend and Moving Average Method is used for the data which exhibit stationary behaviour. In the DSS developed, the above-mentioned methodologies are coded, using Excel VBA programming language, historical data’s behaviour is analysed and forecasts for future spare parts demand are made. The forecasting results are compared based on the minimum error (PAE), to decide upon which is the most appropriate forecasting methodology to use according to the specific spare parts past data behaviour.