BASIC IDEA: Values in more recent time periods have more impact on the forecast and therefore should be given more weight. This method overcomes the drawback of moving averages which assigns equal weights.
Y(t+1) = X S(t) + (1 - X) Y(t)
Y(t+1) = Forecasts sales for next period
Y(t) = Previously forecasted sales of current period
S(t) = Current period sales
X is a smoothing constant with values between 0 and 1.
How to determine the value of X? Use trial and error and choose a number that gives the best forecast.
Finding the best smoothing constant can be difficult.
Usually a poor model for medium or long range forecasts.
Large random fluctuation can severely hurt forecasts. Try double exponential smoothing model here too. The exact formulas for this procedure are given in the references.