Continuous Warehouse Performance Improvement Through Predictive Analytics

Often overlooked amid the talk about the future benefits of IoT is the fact that many current automation advancements remain largely untapped.

Often overlooked amid the talk about the future benefits of the Internet of Things (IoT) is the fact that many current automation advancements remain largely untapped. While many distribution centers do use warehouse management systems to automate order processing and inventory management and warehouse control systems to control movement within the plant, relatively few are taking full advantage of the business intelligence available from the operating data these systems routinely capture. This is partly because warehouse management systems (WMSs) and warehouse control systems (WCSs) typically only interact with messaging sets and don’t share common data structures with each other.  Extracting data from each, analyzing it for patterns and trends and applying any predictive insights gained requires more human resources that most plants can spare.

Enter warehouse execution systems. Warehouse execution systems (WESs) are relatively recent entries in the warehouse automation space. They have evolved to simplify access to and application of operating data. Where a WCS controls the physical plant operations and the WMS controls non-mechanical aspects of the business, such as inventory, delivery and financial operations; the WES provides a central point of analysis for data from both systems. Its purpose is to identify and predict patterns and trends that could optimize operations, and ultimately profitability.

The Role of the WES

In a typical example, an order clerk might receive a customer request for a specific number of products and enter that into the WMS, which then transfers it to the WCS in the form of a pick order. This initiates the process for locating the items, getting them onto a conveyor and delivering them to a shipping area. The WCS will track all activities involved in getting the order to the shipping door, including time spent at each station and activity by operators along the way. If there are system jams and other issues along the way, it tracks those too. But the WCS does not usually map activity to the original order in the WMS.  This is where the WES comes in, helping detecting process anomalies and optimize processes that are only visible when comparing both order and production data.

Detecting Exceptions and Anomalies

 The WES downloads or otherwise connects to the order data in the WMS and the operating data in the WCS.  At the end of this process it aggregates all of that data to confirm that the system or team has picked the right product in the right quantity, packaged it correctly, and diverted to the appropriate outbound lane.  Analyzing the convergence of two previously isolated flows in this way can reveal exceptions and anomalies that might not otherwise manifest until the customer has received the order. By predicting and preventing potential problems before they happen, WES analysis avoids costly after-the- fact troubleshooting and corrective measures.

Process Optimization

In addition to improving customer satisfaction and reducing customer service costs, predictive analytics also optimizes operations. The WMS and WCS collect and store operational data on factors such as energy used, items processed, and operator involvement, and can export that data in SQL or other standard formats. The WES system enables reconciliation and visualization of this data in ways that might indicate future trends affecting key operating variables such equipment effectives, labor requirements, labor requirements, space allocations, workflow and pick times.

Optimizing Equipment Effectiveness

 If a conveyor system was jamming up, for example, the WES could query to see how many jams happen a day, at which times, at which point in the process, which season, and which type of products. Using reporting tools built into the software, it could reveal patterns and trends that could guide improvements in the affected leg of the process and ultimately determination of a Root Cause fix.

Optimizing Labor Requirements

Balancing staff loads must happen in real time based on knowledge of what resources are currently available, where they are needed, and what will be demanded of each, as far into the future as possible. Figure 1 shows a typical labor forecast for the warehouse section of a distribution center. It shows how much work is forecast, how much has been completed and what is left to be done.  This predicts a potential problem situation, in that most units have not completed their assigned projects but have expended the hours allocated. When confronted with such situations, WES software assists with analyses that can help determine the root of the problem.

One large distribution center that was missing an inordinately high volume of promised ship dates, for example, used data from its WES to identify the cause. Analysis of data stored in the system revealed that a number of orders were temporarily misplaced in a large value-added service (VAS) area, where employees inserted flyers and coupons into outgoing orders.  The analysis showed that volume surpassed the number of people available to complete the tasks and orders were falling through the cracks. Modeling the process in the software helped predict the number of additional staff that would be needed to meet ship dates and the problem was solved.

Optimizing Space Allocation

A WES can also help coordinate with other departments to balance space needs in real time. A batch picking operation involving an automated storage and retrieval system, for example, would require a buffer space to receive the order when it arrives. If, however, there is no space available when the picked order arrives, the process would come to a halt. Better analysis of the data, would flag over-commitment of space so that buffer capacity could be increased accordingly to prevent disruption.

Optimizing Workflow

Balancing space, labor and inventory often requires better control over the flow of product among people and places. Figure2, for example, depicts the current performance levels of a distribution center by end-user and work type, revealing where workflow is hanging up or where it is accelerating, enabling process regulation and optimization. From there it is possible to drill deeper to understand what measures might need to be taken to determine the right mix of people or resources.

Reducing Pick Time

WES software enables visualization of relationships among variables with granularity in timeframes not easily attainable by manual analysis. One large department store, for example, had been storing backpacks in the back of the warehouse because they were not typically fast movers.  WES analysis of the data, however, suggested that during the peak season the backpacks should be stored closer to the front where they are more readily accessible to order pickers. While repositioning backpack storage during schools seasons may seem relatively obvious, WES analysis can help determine exactly when and where to move them for maximum return.

Focusing on Future Performance

WES software is most effective when used with ongoing continuous improvement.  In addition to analyzing patterns in relationships among labor, space allocation and workflow data, a continuous improvement program would set and track KPIs in many other areas, including:

  • Alerts
  • Putaways
  • Replenishments
  • Fork work
  • General performance
  • Performance by warehouse
  • Outbound shipment

With so many categories of data, there are numerous possible relationships that can be analyzed together to predict distribution center performance and suggest improvements. Identifying what information you need to be tracking starts with a review of distribution center performance and Identification of the information that decision makers need to improve performance in those areas.  In addition to helping with the heavy data crunching, warehouse execution systems provide guidance via pre-populated report tools based on best practices proven across numerous distribution center automation projects. 

Randy Marble is Vice President of Software Strategies at Daifuku Wynright.

More in Operations