Abstract Fitting an exponential smoothing model to historical data involves two distinct decision processes. First, the presence (or lack) of trend and/or cyclic effects in the data must be identified. Secondly, having identified the terms which should be included in the model, appropriate values for the associated parameters (or smoothing constants) must be chosen. The decision process associated with each of these tasks can be based on a rigorous statistical methodology or less sophisticated intuitive techniques. We have found that the use of microcomputer graphics supported by simple statistical measures gives significant insight into identifying the presence of trend and/or seasonal cycles and measuring the effect of the values assigned to the exponential smoothing constants. This is particularly true when the data set being used has relatively few points. This paper presents forecasting software developed by the authors and illustrates the use of microcomputer graphics in fitting exponential smoothing models to historical data.