Like the previous 2 methods, naïve and moving averages, the exponential smoothing method is going to use historical data of the demand to forecast. However, though this demand forecasting method is a little more complicated, it has some distinct advantages.
One of the benefits of this model is that it takes the most recent observations into account and weights them accordingly. For example, if we are looking at 4 years of demand forecasts, April/May/June of 2014 will likely be weighted differently in April/May/June of 2019. The exponential smoothing method takes this into account and allows for us to plan inventory more efficiently on a more relevant basis of recent data.
Another benefit is that spikes in the data aren’t quite as detrimental to the forecast as previous methods. The most recent forecast has the highest weight and therefore should be the most accurate in predicting demand, as opposed to the moving averages method where the weight for each period is fixed.
While these advantages are great when planning inventory for a specific month, the exponential smoothing model limits our ability to forecast demand using seasonality. This model also cannot be used to identify trends though the next article entitled “trend projection” will resolve that issue.
Though these issues are inconvenient, this model helps us establish a nice foundation for more complex demand forecasting models. Let’s get into it!
How does it work?
To understand this forecasting model, I’ve taken our 4-year historical data set and created a table.
Because this method requires demand planning expertise, you, as the demand planner, are going to have to choose a value for what is known as the “smoothing constant” This constant will determine the impact prior observations have on the forecast. We will represent the constant as α in our forecasting equation.
The equation for exponential smoothing is:
Forecast of period 1 + α*(Actual Sales for period 1 – Forecast for period 1)
The best way to identify your smoothing constant is by understand the difference between a high decimal and low decimal. The smoothing constant is going to be a number between 0 and 1.
The higher a smoothing constant, the more sensitive your demand forecast. This means you will see large spikes of data. This is what a smoothing constant of 0.8 would look like with our data:
The lower a smoothing constant, the less sensitive the forecast and thus the less spikes in demand the forecast will have. This is what a smoothing constant of 0.2 would look like with our data:
Knowing what smoothing constant to use is an important part of demand planning. You need to know your company, know your products, and know your inventory.
If you’re just starting out, there is a method you can use to determine a smoothing constant. That method is trial and error. I would suggest creating 4 columns of possible constants, next to your original. From there you can have a healthy range, let’s just say 0.2,0.4, 0.6, and 0.8. You can then view them on a graph and determine which constant, or which combination of constants best fits the graph.
How it applies to inventory planning?
Forecasting using a combination of quantitative methods and your own experience as a demand planner is what will set you apart in your industry. If you know that your product typically tends to have a high level of noise, then you should plan use a small constant. If you know your product has a low level of noise, it would make sense to use a higher constant because you won’t have to worry about data spikes all too often. Understand these methods will help you keep track of your inventory and make the right decision when planning orders.