The purpose of this dissertation is to propose two new methods for detection of biases in demand forecasting. These methods are adaptations of two statistical process control techniques, the EWMA control chart and the CUSUM control chart (or CUSUM algorithm), to the context of the detection of biases in demand forecasting. The performance of the proposed methods was analyzed by simulation, for several magnitudes of changes in the trend of the series (change from a level series to a series with a trend, changes in the trend parameter, and stabilization of a series with a trend in a constant average level) and with different parameters for all methods. The study was limited to non-seasonal models and to the methods of simple exponential smoothing and Holt’s Exponential Smoothing. The results have shown the superiority of the EWMA method proposed and indicate issues for future research.