To govern is to predict.
But predict is not easy, especially in the supply chain world.
If some algorithms have clearly proved their reliability, the quality of a forecast strongly depends on the demand history characteristics.
To determine the characteristics of a demand history, two coefficients are used:
- The first coefficient is the Average Demand Interval (ADI), it measures the regularity of a demand in time by computing the average interval between two demands.
- The second coefficient is the square of the Coefficient of Variation (CV²), it measures the variation in the demand quantities.
Based on these 2 dimensions, the literature classifies the demand profiles into 4 different categories:
- Smooth demand (ADI < 1.32 and CV² < 0.49):
The demand is very regular in time and in quantity. It is therefore easy to forecast and a low forecast error can be achieved.
- Intermittent demand (ADI < 1.32 and CV² >= 0.49):
The demand history shows very little variation in demand quantity but a high variation in the interval between two demands. Specific forecast methods exist, but the forecast errors will be considerably higher.
- Erratic demand (ADI >= 1.32 and CV² < 0.49):
The demand has regular occurrences in time with high quantity variations. Specific forecast methods exist, but the forecast errors will be considerably higher.
- Lumpy demand (ADI >= 1.32 and CV² >= 0.49):
The demand is characterized by a large variation in the quantity of demand and in the interval between two demands.
This is quite impossible demand to reliable forecast, no matter what forecasting tools and methods are being used.
For all but the “smooth demand” profile, forecast accuracy is not a reliable performance metric. Instead, the planner should think of whether he/she has sufficient inventory planned to satisfy the customer demand when it hits.
A business will typically have parts in each category. Understanding the distribution is an interesting exercise that sets correct expectations on what forecast accuracy is achievable for your business.
Details of the math
To better understand these two coefficients, let’s take an example of a demand history over 12 time buckets:
The ADI is equal to the average interval between two demands:
We can say that, in our data sample, a demand occurs, in average, every 1.83 period.
The CV² is defined as the square of the CV, the coefficient of variation.
To compute the coefficient of variation, we are only considering the non-zero values of the demand history. In our sample, the average quantity equals to 10.57 while the standard deviation equals to 6.43. Therefore, we can deduce the CV²:
If the above sample of 12 periods is categorized as intermittent, it is interesting to compute the ADI and CV² for all the parts of a supply chain and display the results in a scatter plot to identify the general characteristics of a population. Let’s take another data sample for the sake of the exercise:
If summarizing the number of points per demand profile, this would give us:
|Demand Profile||Number of parts||Percentage|
In the above example, 23% of the supply chain parts show a smooth demand history profile, meaning that they should be easy to forecast. 40% of the parts are either erratic or intermittent and forecasting them is already a bit more difficult. For the remaining 36% parts that are lumpy, they are very difficult to forecast.
With this type of distribution, the conclusion is that the management of this company should focus on the safety stock and service levels to cover the variable nature of their demand, rather than aiming at improving the forecast accuracy.