Smoothing gets rid of the noise in your data and it responds well to unexpected trends and seasonality in your business. To learn more about how to use the smoothing technique in forecasting, watch this video.
- I want to show you one more quantitative forecasting technique because it's popular with experienced forecasters, and it gives very good results in most situations. Here's the basic idea. With the weighted moving average technique, you used only a handful of previous data points to calculate an average as our forecast for the next period. The thinking here is that the recent months have the most to offer in terms of market factors.
What if there was a way to capture all the previous months of sales data, but do it in a way that smooths out the big highs and lows in your data? It essentially takes out what we call the noise in your data, and here's the best part. It's easy to calculate. You need only three numbers. Current period sales, current period forecast, and the weighting factor for the current period just like you require with weighted moving average.
With this technique the official name of this number is the smoothing factor. Here's how to calculate it. Take the most recent period's actual sales multiplied by the smoothing factor. Add to it the most recent period's forecast multiplied by one minus the smoothing factor. Where A equals most recent period's actual sales, S equals the smoothing factor represented in decimal form.
Let's use the same weighting from the last example of .40. F equals the most recent period's forecast. In other words, the output of the smoothing calculation from the previous period. That gives you the following. Here is why I like this approach. With the moving average and rollover technique, we're assuming each sales result contains some wisdom about future results. Now we're also including not just past actual results, but all past forecasts.
In other words, we're assuming past forecasts have something to share with us about the future forecast as well. That's pretty cool. Let's calculate the exponential smoothing average in Excel. To follow along, download Exercise File 03_04. In the Excel spreadsheet, we'll start in column J. Now since January is our first month in the data set, we'll take the actual results from January and assume that's the forecast.
So I'll take 4,398 from January and type it into cell J3. To calculate the forecast for February, I'll multiply the previous month's actual results by .4, and then add the previous month's forecast multiplied by .6. Again these numbers are arbitrarily chosen and you can choose whatever numbers work best for you. Next we'll calculate the absolute error by subtracting the actual results by the exponential smoothing forecast.
Once again we'll drag both of these formulas down to the last month. Finally, we'll calculate the average of error to compare to the other forecasting techniques. So let's see how it performed. A mean average error of 1,145. Nice work. Not only is exponential smoothing easy, but it's also great at reacting to new trends that pop up or seasonality in your business. It reacts quickly to changes in the market while still retaining a lot of the wisdom from all past data.
I like that, and you should too. So don't let the name exponential smoothing scare you off. The best forecasters love this technique, because they need just three numbers, and a data set to produce great results.
- Understanding the sales forecasting process
- Defining your market category
- Understanding market dynamics
- Selecting a forecasting technique
- Using quantitative forecasting
- Understand moving averages
- Using qualitative forecasting
- Using estimates from customers, sales reps, and distributors
- Using a panel of experts