Guides

Exponential smoothing in Google Sheets

Exponential smoothing forecasts the next value from the last forecast, corrected by part of its error. Recent data counts most, and older data fades out gradually. Example: the same 18 months of demand used in the moving average guide.

Damping factor and alpha

Textbooks write the method with a smoothing constant alpha (a) between 0 and 1:

Forecast(t) = a * Actual(t - 1) + (1 - a) * Forecast(t - 1)

Excel's Exponential Smoothing dialog asks for a damping factor instead, and the damping factor is 1 - a. StatPak uses the same convention so results match Excel. A damping factor of 0.7 means a = 0.3; the default of 0.3 means a = 0.7. Microsoft's ToolPak documentation suggests smoothing constants of 0.2 to 0.3, which is a damping factor of 0.7 to 0.8.

Run the Exponential Smoothing tool

1. Open Extensions, StatPak, Data Analysis and choose Exponential Smoothing.
2. Input range: B1:B19 with Labels in first row ticked.
3. Damping factor: 0.7.
4. Tick Standard errors, choose Output range C2 and click Run.

Column E shows a second run with damping factor 0.3, for comparison. The column headers give the damping factor.

How to read the output

• Each row holds the forecast for that month, made with data up to the month before. The first row is #N/A because there is no earlier data, and the second row's forecast is simply the first actual value, 120.
• Standard error is the square root of the average squared forecast error over the latest 3 periods, so it starts in row 5. The last one is 17.49096 with damping 0.7 and 9.914795 with damping 0.3.
• Demand here rises steadily, so the heavily damped forecast (0.7) keeps falling behind, and its errors are larger. A lower damping factor reacts faster but also passes more noise into the forecast.

Forecasting the next month

The output stops at the last actual month. Apply the formula once more for July 2026: with damping 0.7, =0.3 * B19 + 0.7 * C19 gives 192.6574.

With formulas

Put the damping factor in a cell, for example G1 = 0.7. Then C3 is =B2, and C4 is =(1 - $G$1) * B3 + $G$1 * C3, filled down. The standard error in D5 is =SQRT(SUMXMY2(B3:B5, C3:C5) / 3). Change G1 to try other values, or use Solver with the Nonlinear method to find the damping factor that minimises the squared errors.

Simple exponential smoothing has no trend or seasonal term, so on trending data it will always lag. For a trend, use a regression on time or a method with a trend component (Holt's method).

Related guides

• Moving average in Google Sheets
• How to run a regression in Google Sheets
• Descriptive statistics in Google Sheets