Exponential smoothing methods provide forecasts using weighted averages of past values of data and forecast errors. They are commonly used in inventory control systems where many items are to be forecasted and low cost is a primary concern. The simplest exponential smoothing method refers to single exponential smoothing (SES), which is suitable for data with no clear trends or seasonal patterns.
For data characterised by trends, Holt’s method is suitable, while for seasonal data, Holt-Winters’ method may be used. Both of these methods require some starting values to feed the algorithm, values which can be obtained by means of backcasting.