Safety stock models enable cover the mistakes present in demand forecast and unexpected variations of this. In this paper is proposed the application of two models of safety stock compared to the currently one in use. These models are applied to five type A references in the distributor, because they represent the most significant sales and greater amount of occupied positions on storage. As a starting point, for the first model is used traditional techniques of forecasting by analyzing the pattern of demand and then takes into account the mean squared error of this prediction and a safety factor to determine the stock. The second model uses a weighted average and the deviation of the time series to find the stock. At the end is presented the performance evaluation ofhe three models and it is shown wich one is the best in having storage savings and guaranteed service level.