Warehouse pick routing is an area many companies and researchers have been working to improve as the e-commerce boom has placed a greater priority on fulfillment speed. Research papers and patents propose solutions for the most efficient ways to move through a warehouse.
DHL reviewed a lot of this research in the process of making IDEA, according to Adrian Kumar, the head of global operations science and analytics at DHL Supply Chain. But much of the research is more theoretical than practical, so DHL worked on a solution that could be deployed by a variety of customers with a variety of picking processes.
“We also leverage some of the knowledge that we have from the transportation side of things where we’re doing routing and scheduling and we’re trying to make the best routes and get the right customers on a specific route,” Kumar said in an interview.
The IDEA system was developed to improve routing for environments where the orders are typically small, but the SKU count in the warehouse is relatively large — a setup that can lead to pickers walking quite a bit.
“A lot of companies … address this issue by adding automation,” Kumar said, pointing out Amazon’s automated warehouses. “But that’s not necessarily for everyone and we cater to a large spectrum of customers. And not all of our customers can justify automation and not all of our customers want that.”
IDEA is meant to provide warehouse locations a way to keep their pick carts if they want to while increasing efficiency. When the orders come into the system, IDEA groups them by location in a process known as clustering.
“If you don’t really have good logic to how to assign orders to a big cart, you’ll end up walking the entire warehouse,” Kumar said.
The software has implemented clustering in legacy-based environments that rely on an ERP and paper-based picking, and in more modern environments where the WMS uses the IDEA output to build pick carts.
When DHL implements the solution for a new customer, it trains the algorithm on that location’s order history, a map of the facility and details on where inventory is stored. And it can be adjusted to look at different factors when making groups. Should it optimize for inventory in the same aisle? Or should it optimize linear distance? The answer will differ depending on the location, customer, width of aisles and other factors.
A WMS without advanced pick logic will do one of two things when orders come in, Kumar said:
Orders are picked on a first-come, first-served basis.
Or the WMS uses first location logic where orders are grouped by their first pick location.
“That could be good logic if all the orders are very large, and all the orders have picks in basically every aisle,” he said, but this logic begins to break down with smaller, more diverse orders.
What IDEA actually looks like to pickers in the warehouse will vary by location, and sometimes they won’t see any difference at all.
“With our integrated version, which is integrated into the tier one WMS, the picker doesn’t see a thing,” Kumar said. “They still get their instructions via a pick assist device” whether its an RF Gun or paper.
In the minimum viable product (the version that works with paper-based legacy systems) the picker will get a pick sheet instead of a set of pick lists. In a legacy system, if 12 orders come in, then the picker will get 12 pick lists.
“We give them a consolidated pick list,” Kumar said. This will group the 12 orders and tell the worker where to pick.
One of the benefits that Kumar sees in this process is the ability to increase efficiency without big hardware changes.
“We’re not changing any existing infrastructure, we’re not ripping and replacing the WMS, we’re just providing a microservice, if you will, to that WMS that allows you to work within existing infrastructure, not put in any new hardware,” Kumar said.