In recent years, along with the development and evolution of technology, the methods for making forecasts have also done so. That is why today there is a third group of forecasting methods based on artificial intelligence in which computation is combined with quantitative and qualitative methods. Even when companies have been using the first ones complemented by the latter to make demand forecasts, these have not been entirely accurate in recent years, causing bad planning, which has led to deficits in market coverage in some cases , excess of inventory in others, leading to losses in sales and important capital in lost profits. Therefore, the objective of this research will be to determine the utility of the method by Neural Networks to forecast the demand of brands of mass-use products with respect to the three main methods used by companies that are dedicated to the distribution of said products in the market. city of Bogota. For its development, an applied research will be carried out with a post-facto design, with a quantitative approach, of an analytical type, in which the three demand forecasting models most used by the city's mass distribution companies and their results will be applied. they will be compared by estimating their mean square errors with the results obtained with a neural network model for a three-year data set provided by one of these companies. The proximity of the results of forecasts, given by the lower mean square error, with respect to the data of what happened in the analyzed years, will allow establishing the usefulness of the method of artificial neural networks for the forecast of demand in mass consumer brands.