Cluster analysis can help determine where to store inventory, reduce split shipments
When a business receives an online order from a customer that contains multiple items in different SKUs, then it must decide how best to process the order and minimize split shipments. A recent article by researchers in China examined using a grouping algorithm with product categories for the purpose of reducing saliva shipments.
An order is considered split when it contains two or more items that are stored in different warehousing locations, resulting in multiple shipments. For the research, they worked with “one of China’s largest online supermarkets” and used company data over a six-day period. During this period, the majority of the company’s daily orders consisted of multiple items. And among multi-item orders, 76% were in multiple categories.
Their goal was to determine which warehouses should stock which categories to ensure that orders were divided as little as possible.
In an online supermarket in China, the orders are mostly multi-item
% of orders received by an online supermarket in a six-day period
Manjeet Singh, director of science and analysis of global operations at DHL, said a company can reduce the number of packages and the overall cost of transport per order if it can reduce split shipments.
âThis is why companies would like to have so many product category items, which are usually ordered together, in one place,â said Singh, who was not involved in the research.
The document also notes that more split shipments increase environmental pollution due to the slight increase in packaging.
The researchers formulated a model that identifies products by SKU, but organizes products by category. The goal was to create an algorithm that would help them distribute these categories to the available warehouses in a way that reduced split shipments.
The algorithm worked to reduce what they called “outbound links”, that is, when two categories of the same order are in different warehouses.
Grouping inventory based on categories rather than SKUs made it easier to solve the problem mathematically. But it also makes sense from a warehouse operations perspective, Singh said.
âYou don’t want different types of items that require a lot of handling or different types of handling in the same warehouse, generally speaking,â he said. “They want similar items in the warehouse because your warehousing costs are going to go down.”
He said that is why Amazon usually stores inventory based on size.
The researchers found that the algorithm they developed was able to reduce outbound links compared to previous methods.
“This is a significant improvement over the actual distribution of categories in practice,” the newspaper read. âThis means that if the online retailer uses our best category breakdown, their number of split orders will drop dramatically. “
Singh said the researchers have developed an “interesting model” and the ability to allow a company to create its own categories can certainly simplify the grouping process. But he stressed that the person creating the categories needs to understand the company, the products and the geography. And this is in line with the work that operations researchers have been carrying out for some time.
“It’s nothing revolutionary … it’s something we did here, we didn’t use the model like this, but we used techniques [to solve the problem] in the same way, âhe said.
The use of clustering as an analytical tool has found a number of uses in warehouses in recent years. Clustering is the ability to determine whether observations, such as orders or product categories, belong to distinct groups.
“Identifying such groups may be of interest because the groups may differ with respect to certain properties of interest, such as spending habits,” it reads. the latest edition of “An Introduction to Statistical Learning.”
Singh used clustering in his research that helped DHL optimize packaging sizes in warehouses and lower shipping costs. It is a technique that can not only be used to determine in which warehouse to store inventory, but in which aisle to store inventory.
He noted that an interesting next step for this research would be to use machine learning to generate the categories rather than relying on a human to generate the groups.
âIf you happen to have a lot of SKUs, a lot of items and you don’t want to leave that to one individualâ¦ it might be worth using a machine learning technique. [to create the categories],” he said.
The researchers had their own ideas for the next steps: to allow one category to be kept in multiple warehouses. Their current model assumes that a product category is simply kept in a warehouse. But allowing the fastest items to be stored in multiple warehouses could reduce splits even further, they suggest.
This story first appeared in our weekly newsletter, Supply Chain Dive: Operations. Register here.