Demand Patterns

In demand planning, one of the biggest challenges is achieving forecast accuracy goals that often seem impossible for certain products. This isn’t always the forecaster’s fault; the issue may lie in the metrics used to measure forecasting accuracy. A key factor here is forecastability, which is often misunderstood or overlooked.

Forecastability: The Key to Accurate Forecasting

Forecasting typically involves gathering historical data, applying forecasting models, and predicting future demand. However, the reliability of this process depends heavily on the nature of the product’s demand pattern. Some products have predictable, stable demand, while others show high variability, making accurate forecasting much more difficult.

Forecastability can be better understood through two key coefficients:

• Average Demand Interval (ADI): This measures the regularity of demand over time by calculating the average interval between two consecutive demands.

• Coefficient of Variation (CV²): This measures the variability in demand quantities. A high CV² indicates large fluctuations in demand, while a low CV² signals stable demand.

Together, these two coefficients help us classify demand profiles into four distinct categories, which influence forecasting methods and accuracy.

Parameters calculation

• The Average Demand Interval (ADI): It measures the demand regularity in time by computing the average interval between two demands.

Ti: Period between two consecutive demand periods
N: Number of all demand periods.

• The Square of the Coefficient of Variation (CV²): It measures the variation in quantities.

The threshold values of ADI and CV² are 1.32 and 0.49, respectively

Based on these 2 parameters the demand classifies the demand profiles into 4 categories:

• Smooth:

ADI < 1.32 & CV² < 0.49
Characteristics: Regular demand both in timing (consistent intervals between demands) and quantity (steady sales).
Forecasting Ease: High predictability, easy to forecast with low error levels. These products are predictable and can be sold consistently, such as daily or weekly.

• Intermittent:

ADI >= 1.32 and CV² < 0.49
Characteristics: Irregular timing of demand, but with consistent quantities when demand occurs. For example, a product may sell this week but not for several weeks afterward.
Forecasting Ease: Moderate difficulty. Although some forecasting methods can handle intermittent demand, irregular intervals make accurate predictions more challenging, resulting in higher forecast errors.

• Erratic:

ADI < 1.32 and CV² >= 0.49
Characteristics: Regular timing of demand, but with large fluctuations in demand quantity. For instance, a product may be sold daily, but one day it may sell 5 units, another day it could sell 110 units.
Forecasting Ease: Difficult to forecast due to high variability in the quantity sold, even though the demand occurs at regular intervals. Predicting the exact demand is challenging due to erratic sales volumes.

• Lumpy:

ADI >= 1.32 and CV² >= 0.49
Characteristics: Both irregular demand intervals and high fluctuations in the quantity sold. This type of demand is unpredictable, with large gaps between sales and varying quantities when sales do occur.
Forecasting Ease: Very difficult to forecast. Lumpy demand is essentially unforecastable due to its high variability in both time and quantity. A common solution is to hold safety stock to mitigate the risk of stockouts.

References
Zhuang, Z., Yu, Y., Chen, A., 2022. Data Science and Management
https://frepple.com/blog/demand-classification/