A system of drawing up medium-term sales forecasts has been developed for the range of a large concern in the food industry and is intended to form part of the integrated operations planning in the logistic sector. The forecasts in this system are based on a break-down of sales volume into the following components: - level of sales; - trend; - seasonal influence; - special influences (promotions, price changes, movable feasts, etc.). Basic values for the various components are determined from past values. The forecasts of weekly sales in the stipulated period (6 months) are based on these values. The basic values are adjusted from week to week as new historical sales data become available. Basically, the exponential smoothing method is used to determine sales level, trend, and factors for seasonal and special influences. A pragmatic procedure is used to adapt the smoothing constants. The most important aim was a complete forecasting system. The system consists of a series of computer programmes which at present are run on an IBM/360-50. It contains an analysis of past data, the forecast and the up-dating of data. The output is such that the person responsible, the product manager, is able to control and correct the system. We are fully aware that this system and the methods being applied are capable of being improved. Apart from this, however, we believe that there are several basic difficulties in drawing up sales forecasts which will hardly be obviated by means of improved systems or methods. In particular, there are the well known problems connected with the quantitative forecasting (quantification) and the temporal forecasting of the various factors affecting sales. The quality of the forecasting system presented here was examined in a test involving selected products over the whole of 1968.