By James Bell
We go through three different types of simple forecasting. The purpose here is to introduce you to basic forecasting and budgeting techniques. You can use this to predict the demand of a product or how much of something to make. You can also use it to create budgets and more. There are a lot of other forecasting methods and even more advanced time series forecasting methods.
The importance of forecasting well cannot be stressed enough. Resource allocation is crucial to a successful business. Accurate forecasts can improve your relationship with your investors, customers, suppliers, and business partners. Your ability to better allocate resources can be a source of competitive advantage especially in saturated markets and product positions.
We go over multiple methods here. The obvious question that comes up with multiple options is which one to use. While this seems like a simple question the answer can be as complex or simple as you want it to be.
If you do decide on one method, we suggest “checking in” to make sure that you are still using the most accurate method as time goes on. We will go over this in a future article. As things change over time, you find find that one method that has worked well in the past may not work as well going forward.
We will explain variables in each section and introduce new ones as we go.
A simple moving average uses prior periods to forecast future periods. It’s the simplest method of the three that we go over. You will need some previous data in order to use this method.
We take the total sum of the amounts from previous periods and divide them by the number of periods.
where
Forecast
Actual Historical Values
Number of Periods
This is the amount that we are forecasting. For example, we could take the previous 3 weeks and average them to predict the 4th week.
These are not our previous forecasts, but actual historical data for each period 1, 2, and 3. This data has to exist to use this method.
This is the count of how many numbers you are adding together. Think of this as a basic average calculation.
This is similar to the simple moving average but we add weights to each period. As business, markets, and economies change over time, you may find it more relevant to weight current periods heavier than past periods. Including prior periods is a sort of mean reversion that helps smooth things out over time.
where
Weights 1, 2, and 3
The total of all weights must equal 1. In the previous method, we essentially assigned equal weights to each period. So if you equally distribute the weighting, this forecast will match the previous example.
In reality, our most recent period may actually be more relevant than prior ones. Depending on how strongly you feel about this, you can increase the weights of the most current actual to calculate a forecast that may be more accurate. I often see the most current period weighted 0.5, then 0.3 for the one prior, and then 0.2 for the earliest period.
This is a more advanced version of Weighted Moving Average. Exponential Smoothing is great when you don’t have a lot of data to begin with.
This method is more consistent from period to period. There is a lag that is built into this method that is based off of changes in the underlying average. As it changes, the forecasting will change and you will see that it has a smoothing behavior which is how this method gets it’s name.
where
Forecast
Smoothing Parameter
Actual for Current Period
Forecast of Current Period
While the subscript looks a bit different, it functions the same as the others. This is the forecast for the next period.
Alpha is the weight that we are using to smooth out our forecast. The formula is set up so that alpha must be between 1 and 0 as the total weight must equal 100% just as it is in other weighted average type formulas. A higher alpha means a higher weight for the most recent actual. This reduces the smoothing effect. Inversely, a lower alpha has a greater smoothing affect as more lag is introduced.
This is the actual historical amount for the current or most recent period
This is the amount that we forecasted for the current or most recent period.
In later articles, we are going to take these basics and add Trend-Adjustments through Linear Regression Analysis. Here is the article where we cover forecasting errors and bias. This helps us make the decision of which method to use as the method that produces the least variance or bias is most desirable.
15 March 2020
15 December 2019
14 November 2019
07 November 2019
12 October 2019
26 September 2019
11 September 2019
Powered By Impressive Business WordPress Theme