Does demand forecasting matter to retailing?

Time series forecasting in retail plays an important role for retailers due to its capacity to aid managers in strategic planning to meet demand. In this article, a comparative experiment between two demand forecasting models is conducted to forecast an aggregate time series in Brazilian food retail on a twelve-year horizon. Performance was analyzed by comparing the accuracy of a linear model, Seasonal Autoregressive Integrated Moving Average (SARIMA), and a non-linear model, and Wavelets Neural Network (WNN). Performance was validated by the Mean Absolute Percentage Error (MAPE) and U-Theil, with a view to achieving the most accurate measurement of demand responsiveness, using the Fill Rate (FR). WNN had the best performance, with a FR of 98.43%, whereas SARIMA achieved a FR of 102.23%. Demand forecasting using WNN can be considered a resource for differentiation between retail organizations when it is incorporated into their operations. This article makes three contributions to the field with an application to retail and strengthening the argument for choosing a promising technique that can be used by managers to aid their decision making.