Reducing the total travelling distance of order picking in a warehouse by introducing class-based storage

This research is conducted at a 3PL company and provides insight into reducing the travelling distance of order picking by introducing a Class-Based Storage policy. Instead of storing items randomly in a zone, I propose to classify SKUs based on their average length of stay into 2 (fast- or not-fast-moving) or 3 classes (fast- , slow- or non-moving). Because the SKU assortment is constantly changing, I use the Random Forest classification algorithm to classify SKUs based on SKUs with similar features. By storing fast-moving items closer to the depot, the total travelling distance can approximately be reduced by 4.2% when classifying into 3 classes. If the classification is performed perfectly, a reduction of 20.4% can be expected. In the case of 2-class classification, the travelling distance can approximately reduce by 3.2%, while 17.2% is achievable with perfect classification. So classifying into 3 classes outperforms classifying into 2 classes, but is also more complex to implement.