Stop Using Segmented Forecasting
With advancements in technology and data analytics, demand forecasting is experiencing a significant transformation. The traditional, widely used approach of first segmenting products and then applying distinct forecasting techniques to each segment may no longer be the most effective or accurate method. This article first discusses the drawbacks of employing segmentation, then navigates towards alternatives (optimization engines and machine learning) that promise substantial enhancements to your forecasting strategy.
Demand forecasting has always been crucial for supply chain management. It helps businesses to plan and manage their resources effectively and avoid unnecessary costs (we recently demonstrated that there is a direct correlation between forecasting accuracy and business value). However, as market dynamics are complex, we need forecasting methods that can capture this complexity and deliver accurate predictions. In this context, we delve into two alternatives to segmented forecasting: SKU-by-SKU optimization and global machine learning models. We will discuss their benefits and address the common concerns about their applications.
Understanding Segmentation in Demand Forecasting
What is Segmentation and Why is it Used?
Segmentation is the practice of classifying elements into clusters based on common attributes or patterns. Supply chain practitioners often use this method to organize products into distinct categories.
SKUs are commonly grouped based on characteristics like their historical sales volume, volatility, or profitability — think traditional ABC or ABC XYZ segmentation techniques. Classifying your products is not a bad idea per se, but you need to pay attention to many details: how do you classify them, when do you review this classification, and what do you use these classifications for? Even though you can use these segmentations for many…