A New Demand Forecasting Framework Based on Reported Customer Forecasts and Historical Data

The main purpose of the study is to design a forecasting framework through a better understanding of the structure of the customers’ own forecasts to minimize the forecast error. Different than the conventional forecasting methods, our framework has an extra input, which is the set of reported customer forecasts; which is mostly based on customers’ own production plans and capacities. Furthermore, our framework also takes the advantage of the classical methods: Winter’s Method in its pure form and Winter’s Method merged with Decomposition Method are taken into consideration, because these methods are able to handle stability, seasonality and trend in the demand data structures. The general proposed model is designed to embed the customer forecasts to the forecasting framework with a scientific methodology. A new method is created with a logic of a calculation of reliability indices for each customer and product pairs with following some statistical procedures. As a result of the method which can calculate the reliability of each customer, a forecast output is provided. On the other hand, forecast Combination Method is adjusted with respect to the nature of the problem to combine the output of Winter’s Method’s and the output of Customer Forecasts Method. In order to implement all these methods, the study is practiced in a well-known, international energy company. The demand data are divided into such subgroups as; Regular, High Variance, New Born, Rare, On Off and Inactive based on the demand structure of the data, to observe how these subgroups would react to the method. In order to achieve all these routines of the forecasting framework, we design a Decision Support System in Excel Visual Basic Applications (VBA).