Previsões de demanda em uma indústria de manufatura terceirizada do ramo de eletrônicos

Abstract

Accurately forecasting demand is a difficult task, even when you have the appropriate information to help with this work. When a company's demand is dependent on information coming from a customer, this task becomes even more difficult, thus harming all the planning of raw material acquisition and factory capacity to meet demand. One of the lines of action that suffer the effects of working with customer-dependent demand is called Contract Manufacturing, that is, companies that produce electronics in a third-party manner. Thus, this study aimed to study the expected demand with the customer, with the aim of finding improvement points that will help the client to improve his forecasts. For this, this work approaches the concepts of demand forecasting, exponential smoothing methods and ARIMA methodology, as well as prediction combining methods. From the concepts discussed, the same were applied with the help of Statgraphics software to define the models that best fit the historical series of data. A comparison was made between the combined demand forecasting model and the forecast model used by the client, indicating which is the most appropriate for each historical series of data. In some cases, the combined model achieved MAPE 66% lower than the method used by the client.