Time-Series Forecasting: Exponential Smoothing Part 1

We have already looked at the standard moving average and the weighted moving average. However, there is another forecasting tool that an analyst or project manager can use. It is both simple and sophisticated. It is known as Exponential Smoothing.

Using only the most recent data, you can create a forecast for the next month's demand or sales. This method places an exponentially decreasing weight on each sequential piece of older data. With a series of data, you can see how the exponential smoothing really "smoothes" the results so that there are fewer noticeable spikes and dips.

The formula for smoothing is:

Now the tricky part is determining what the smoothing constant (alpha) should be. A table could be a handy tool to help. The constant used for the most recent period would be just alpha. NOTE: I'm going to use a little "a" to represent alpha, because I'm not sure how to make that symbol out of HTML.
For the 2nd most recent period, your weight should be: a(1-a).
Then for the third most recent period, the weight would be: a(1-a)^2.
The fourth most recent period would have a weight of: a(1-a)^3.
The weights would continue like this.
All the weighted values are then summed to give you a forecast.

I know it's a little confusing. However, in my next post in this forecasting series I plan to do a sample problem with exponential smoothing. That may make it easier to understand. If somebody reading this has a better explanation of exponential smoothing, drop me a comment. Your contribution would be welcomed.

Thanks for visiting. -------Sincerely, Trevor Stasik.

---

To return to initial post about forecasting, click HERE.
To view the next post in my forecasting series, click HERE.

---

exponential smoothing, forecasting, financial tool, demand forecast