What’s an inventory forecast?
Inventory forecasting helps estimate future product demand over short, mid, or long-term periods. It is a core part of any inventory planning strategy. By predicting demand in advance, you can improve customer service, inventory control, and capacity management.
You can also forecast cost-related data, such as the expected cost when placing future orders. Lead time is another key factor that can be forecasted, although it tends to get shorter over time. No matter what you’re forecasting, keep these principles in mind:
- Forecasts are more accurate in the short term. Accuracy drops the further you project.
- Forecasts will contain errors. Understanding the error percentage is crucial.
- Forecasts are not a replacement for actual demand data.
Types of forecasts
There are many ways to forecast demand, and no single method is always the best. The right approach depends on the time frame and the available data. Forecasts generally fall into three time-based categories:
- Long-term forecasts: several years into the future
- Medium-term forecasts: three months to one year ahead
- Short-term forecasts: a few weeks ahead
Forecasts can be data-driven or based on expert judgment. If historical demand data is available and reliable, it becomes the preferred source.
Forecasting techniques
We previously explored the Economic Order Quantity (EOQ) method to determine how much to buy. In this post, we’ll look at three basic forecasting techniques that rely on past demand data.
For illustration, we’ll use sales data from October, November, and December 2022 to forecast the next three months of 2023.
Moving Average
This method calculates the simple average of a set number of past periods to predict the next one.
| October 2022 | 114 |
| November 2022 | 119 |
| December 2022 | 137 |
| January 2023 | ? 114 + 119 + 137 = 370, 370 / 3 = 123 |
| February 2023 | ? 119 + 137 + 123 = 379, 379 / 3 = 126 |
| March 2023 | ? 137 + 123 + 126 = 386, 386 / 3 = 129 |
Weighted Moving Average
This method applies different weights to each period, with the most recent data usually getting the highest weight. It’s useful for short-term forecasts of stable products.
Note: The weights must total 1.00. In this example, we’ll use 0.6 (most recent), 0.3, and 0.1
| October 2022 | 114 |
| November 2022 | 119 |
| December 2022 | 137 |
| January 2023 | ? 137 * 0.6 + 119 * 0.3 + 114 * 0.1 = 129 |
| February 2023 | ? 129 * 0.6 + 137 * 0.3 + 119 * 0.1 = 131 |
| March 2023 | ? 131 * 0.6 + 129 * 0.3 + 137 * 0.1 = 130 |
Linear Smoothing
Similar to the weighted moving average, but the weights decline linearly using a formula. It works well for short-term forecasts of stable products.
Weight for one period prior = 3 / ((n² + n)/2) = 3 / 6 = 0.5
Weight for two periods prior = 2 / 6 = 0.3
Weight for three periods prior = 1 / 6 ≈ 0.17
| October 2022 | 114 |
| November 2022 | 119 |
| December 2022 | 137 |
| January 2023 | ? 137 * 0.5 + 119 * 0.3 + 114 * 0.17 = 126 |
| February 2023 | ? 126 * 0.5 + 137 * 0.3 + 119 * 0.17 = 128 |
| March 2023 | ? 128 * 0.5 + 126 * 0.3 + 137 * 0.17 = 129 |
Conclusion
Forecasts rely on data and assumptions. They are not perfect and will contain errors, whether from the model, data, or interpretation. Still, a well-informed estimate is far more useful than no estimate at all.
In future posts, we’ll explore more advanced methods like exponential smoothing and least squares regression, and how these can be automated using inventory management software.