Regular seasonal patterns appear in most business data. The weather affects the sales of everything from bikinis to snowmobiles. Around holiday periods, we see increases in the number of retail sales, long distance telephone calls, and gasoline consumption. Business policy can cause seasonal patterns in sales. Many companies run annual dealer promotions which cause peaks in sales. Other companies depress sales temporarily by shutting down plants for annual vacation periods.

Usually seasonality is obvious but there are times when it is not. Two questions should be asked when there is doubt about seasonality. First, are the peaks and troughs consistent? That is, do the high and low points of the pattern occur in about the same periods (week, month, or quarter) each year? Second, is there an explanation for the seasonal pattern? The most common reasons for seasonality are weather and holidays, although company policy such as annual sales promotions may be a factor. If the answer to either of these questions is no, seasonality should not be used in the forecasts. 

Our approach to forecasting seasonal data is based on the classical decomposition method developed by economists in the nineteenth century. Decomposition means separation of the time series into its component parts. A complete decomposition separates the time series into four components: seasonality, trend, cycle, and randomness. The cycle is a long range pattern related to the growth and decline of industries or the economy as a whole.

Two worksheets are available for seasonal adjustment. MULTIMON uses the ratio-to-moving average method to adjust monthly data. ADDITMON uses a similar method called the difference-to-moving average method to adjust monthly data. It may be necessary to test both of these worksheets before choosing a seasonal pattern.