Time series prediction for economic processes is a topic of increasing interest. In order to reduce stock-keeping costs, an appropriate forecast of the demand in the future is necessary. In this paper we use neural networks for a short term forecast for the sales of articles in supermarkets. Multilayer perceptrons are trained on the known sale volume of the past for a certain group of related products. Additional information like changing prices and advertising campaigns are also given to the net to improve the prediction quality. The net is trained on a window of inputs describing a fixed set of recent past states by the back-propagation algorithm [1]. For enhancement the algorithm is also implemented on parallel systems.