The smoothing constant determines the level at which previous observations influence the forecast.

- Large weights result in faster changes in the fitted line; small weights result in slower changes in the fitted line. Therefore, the larger the weights the more the smoothed values follow the data; the smaller the weights the smoother the pattern in the smoothed values. Thus, small weights are usually recommended for a series with a high noise level around the signal pattern. Large weights are usually recommended for a series with a small noise level around the pattern.
- A different way of choosing the smoothing constant: for each value of α, a set of forecasts is generated using the appropriate smoothing procedure. These forecasts are compared with the actual observations in the time series and the value of a that gives the smallest sum of squared forecast errors is chosen. (The commonly used values for a are between 0 and 1, with steps of 0.2 or less.)