Representative End User Clients
Representative Automation Clients
Representative Software Clients
Summary
Manhattan Associates, a leading provider of supply chain software solutions, held its Momentum 2018 conference May 21 through 24 in Hollywood, Florida. At the conference, ARC Advisory Group learned that Manhattan has made some interesting and significant enhancements to its warehouse management system (WMS):
- The enhanced solution uses machine learning and advanced optimization techniques to improve productivity, particularly e-commerce order fulfillment
- The enhanced solution simplifies the automation hierarchy
Variety of Systems Used to Support Warehouse Productivity
A little background is needed to fully understand the significance of the enhancements announced at the conference. A warehouse management system allows a company to accurately fulfill orders while improving labor productivity. Labor productivity was driven by grouping orders into waves. A wave might be all the orders that will go out on the UPS truck scheduled at 10:00 am, or all the orders going to stores one through five that will ship on a truck at dock six. These waves of work facilitated the intelligent grouping of orders fulfilled by a particular associate, which reduced travel time.
A WMS also contains logic for receiving goods, dock appointments, moving goods from storage to forward pick locations, value-added services (like kitting), and all other end-to-end processes associated with warehousing.
When warehouses have advanced automation - conveyors, high speed sortation, automatic storage and retrieval systems, and so forth – companies generally implement both a WMS and a warehouse control system (WCS). The WMS contains the order fulfillment logic, the WCS the MOVE logic. MOVE logic looks to manage how a product is transported on automated systems from point A – for example, a pick location – to point B, perhaps a pack station or a shipping dock.
But over time, with the growth of e-commerce, another three-letter acronym software solution was born – the WES (warehouse execution system). A WES generally includes both the MOVE logic from a WCS and some of the WMS’ picking logic.
Some of the leading WMS solution providers had batch picking logic, but lacked the logic to optimally pick e-commerce orders. E-commerce orders are typically small, one or two items. And rather than batching these orders to gain labor efficiency, it often makes sense to drop these orders - or at least the orders that customers were promised would be delivered in a day or two – to the floor as soon as they are received. Thus, e-commerce orders are fulfilled with something known as “order streaming” logic.
WMS suppliers responded slowly to the growing need for optimized e-commerce picking, creating the opportunity for WCS suppliers to reach up and grab this space. But it never made sense for “order streaming” logic to be supplied by WCS (rebranded as WES) suppliers, because order streaming is only one part of the end-to-end processes that need to be supported in a warehouse. Further, many warehouses need to perform both traditional and e-commerce fulfillment, and WES solutions lacked the wave logic needed to optimally support traditional pallet and case fulfillment.
Order Streaming on Steroids
Manhattan Associates first announced that it would enhance its WMS with order streaming at last year’s Momentum. At this year’s conference, the company announced a solution that might best be described as “order streaming on steroids.”
In orchestrating when e-commerce orders are dropped to the floor, WMS solutions have mostly operated with simple rules. For example, if an order needs to be delivered to a customer tomorrow, “drop it immediately.” If it does not need to be fulfilled for three days, “wait two days before dropping it to the floor.”
What Manhattan has done is to greatly enhance the orchestration logic with real-time awareness of the capacity of both men and machines, as well as the availability of the necessary inventory at the right location, to do a piece of work at a given time. Rush orders get dropped first, but better awareness that the system still has capacity available to fulfill orders that might not need to be shipped for a few days, means both equipment and people can be more fully utilized.
Fully understanding the capacity of the system to do work, in turn, means that the system must understand at a very granular level how long it takes to do a given piece of work. When it comes to how long it takes a person to do a given task, Manhattan Associates has a labor management system (LMS) that already contains this logic. An LMS uses precisely determined standards to understand exactly how long a task should take based upon the distance to be traveled, where an item is placed on a shelf, and the size and shape of the item to be picked. But implementing an LMS requires money and effort, and an LMS often does not include all the time standards logic needed to fulfill orders that involve material handling equipment.
Manhattan is looking to tackle these issues with machine learning. As workers or machines fulfill an order, the system will learn over time how long the task should take. Adam Kline, a Senior Director of Product Management at Manhattan Associates, points out that machine learning “remains important even if a customer has LMS. Labor can tell us two things. How long a task should take and how long it did take. The ‘should’ comes from engineered standards. Conditions change, however, so machine learning can offer a perspective on how long the system anticipates the task will actually take given the combination of historic task data and additional conditions such as item characteristics, number of stops on the task, location information, day of week, and hour of day.”
Over time, as the capacity issues are understood, the system can use mixed-integer optimization – the kind used in advanced production scheduling systems – to understand exactly how many orders can be dropped to the floor in the form of executable tasks. While generating the tasks, the system leverages another optimization methodology called Adaptive Large Neighborhood Search (ALNS), which utilizes a robust distance calculation framework provided by the new Warehouse Map component.
Mr. Kline summed it up by explaining, “Execution is constantly monitored by the planning component with the help of real-time signals on task statuses, worker availability, processing speed, and resource utilization. This constant feedback is blended with carrier cut-off requirements and initial priority of orders. Then, heuristics (rules-based systems) are used to re-prioritize which orders are more and less important at a given point in time.” It may be that out of 500 potential tasks that could be worked on, based on the priority of the tasks and the capacity available in the system, only 50 are dropped to the floor. The other 450 orders are sent back into the order funnel engine for replanning once capacity becomes available. “Any new order arrival, or change in the execution state and performance, are factored into the new plan which results in an adaptive work planning process.”
Conclusion
ARC believes that this is a significant enhancement employing advanced optimization techniques not seen in competing WMS solutions. However, since the addition of the term “WES” to the lexicon tends to confuse the entire discussion, it may be in the industry’s best interest to avoid using it unless absolutely necessary.
ARC Advisory Group clients can view the complete report at ARC Main Client Portal or at ARC Office 365 Client Portal
If you would like to buy this report or obtain information about how to become a client, please Contact Us
Keywords: Warehouse Management, Automation, Optimization, Machine Learning, ARC Advisory Group.