Prescriptive Analytics in the Warehouse: How Technology Enables Execution

prescriptive analytics

3PLS AND THEIR clients are experiencing the brunt of an unprecedented level of growth in online orders due to the pandemic. Whenever you walk into a Distribution Center (DC) nowadays, there’s a high probability that the team is behind schedule, product is being cut from customer orders, and there’s not enough labor on the floor to make ends meet. It’s likely to get worse.

Pitney Bowes’ State of the Consumer survey found that 54% of consumers shop online more now than they did prior to the pandemic. This growth in demand and change in shipping medium is making 3PLs scurry to meet fulfillment times and order requirements.

While the profits for 3PLs may be great, numerous challenges abound due to the pandemic, including:

  • Legacy system integration with newer technologies
  • Need to share a single source of the truth across all parties involved
  • Labor availability and quality of labor
  • Struggle with lack of real-time measurements
  • Capacity challenges in transportation
  • Customer behavior has shifted
  • Push to become a more planning-centric operation, rather than a reactive operation

Being able to fulfill orders quickly and accurately is the main goal of any 3PL in the current COVID-19 marketplace. This takes innovative technology that turns the 3PL into a provider of choice. You can’t do it by just throwing technology into the wind. It’s a journey 3PLs must undertake to build a digital strategy.

To effectively manage a site, the dark planning era has arrived. It is imperative that companies automate all operational warehouse decision-making in a capacity-considerate way, freeing up people to fight the fires that pop up. An autopilot planning system leads to indirect labor savings, increased fill rate, increased load on time, and productivity improvements across direct labor by optimizing the work historically released and executed by warehouse operators.

wms systems

Steps to Building a Digital Strategy

Step 1: Acquire data about the operations. Some 3PLs use spreadsheets to manage the data flow, but best-in-class 3PLs automate data collection and management with a variety of tools. Once you acquire the data, you need to understand how to pull all this information into one place. You need to prioritize inbound and outbound processes as the warehouse has only so much space for inventory, a specific number of dock doors, and a limited number of workers and time to fill orders.

Most DCs already have a warehouse management system (WMS) in place but struggle to see the value in upgrading it. A WMS is stellar at ensuring execution happens, but terrible at orchestrating activities. You need to balance who should be doing what. It is imperative to understand who is working in each zone and what work is being done in those zones.

WMS systems are good at managing inventory age and space; managing a work queue; zoning a DC; and effectively cataloging and visualizing all receipt, shipment, and order information. Areas where a WMS can fall short, however, include:

  • Understanding labor, space, and task constraints to optimize full-site operations
  • Intelligently allocating inventory and managing cuts
  • Creating chained tasks that factor in proximity-to-task for all associates
  • Managing receipts and shipments simultaneously to ensure the right inventory movements are optimized
  • Transferring inventory across multi-building campuses

The WMS provides critical data used throughout the warehouse, but this data needs to be visible and usable. To make data more accessible, many companies are adopting a new breed of technology known as a “data replica,” which creates an easy-to-access database that pulls in real-time WMS data for querying, visualization, and alerting. More and more vendors are appearing in this space every quarter.

Step 2: Predict and identify challenges in the DC: How can prediction add value? It helps you to know when you are short of inventory and to understand capacity challenges. Planners need to know where problems will occur in the warehouse. How do you decide which trailer to load next? The WMS says that 7:00 a.m. is the earliest appointment time, so you select the next trailer to fill. But what if you don’t have the inventory to fill that trailer? What do you do?

You want to use the data to understand the future state of the DC. This is often called “what-if” planning and leverages digital twin technology to perform this task. A digital twin is a mathematical model of the warehouse that analyzes all future-facing activities to predict what is likely to happen in the future. An excellent digital twin will account for labor, shipments, inventory availability, tasking, and space/resources.

Step 3: Prescribe the optimal workflow for operations. Prescriptive analytics uses constraint-based mathematics to build optimal plans. You work backwards from loading and work in parallel with other tasks. When paired with a digital twin, constraint-based optimization technology can prescribe a sequence of events, create a feasible schedule, and minimize touches and labor.

The Road Map – Dark Planning

So why do we want a prescriptive warehouse, and how do we make it happen? The core objective is to maximize the customer-facing output of the DC while understanding and respecting all of the different space, labor, and process constraints that exist within the building. This reduces touches, cuts travel, and increases the capacity per labor unit to drive value. However, it does require buy-in from a few key individuals. To convince planners of the value “Dark Planning” can provide, there are three key stages that each site needs to work through.

1) Introduce a manual side-by-side comparison to build comfort. Data is pulled from core execution systems into a transactional data service. Planning solutions access the data from the service and create an optimized plan every hour. Planners review key planning reports to determine a best course of action. The planner manually allocates inventory and modifies work queues in the WMS based on the planning solutions recommendations.

2) Integrate a semi-automated workflow that requires planner approval. The planner reviews new planning solution recommendations and determines which ones to approve/enable. This information is persisted inside of the planning solution, and approvals are picked up by the WMS integration services and automatically injected into the WMS.

3) Go Dark. The planning system recommendations are picked up by the WMS integration services and automatically injected into the WMS—any changes from run-to-run are persisted and updated in the WMS. The planner reviews the plans on their own schedule and tinkers with them when they get bored.

This orchestrated Dark Planning accounts for all functions (replenishment, cartonization, loading and picking), actors (conveyors, pickers, loaders, and mobile systems), and constraints inside of the DC to map out the ideal plan for execution as conditions continue to change. Instead of fighting fires all of the time, planners can start to focus on exception handling. Dark Planning pushes us into a future era where DC’s are proactive about their work queue, and able to communicate exactly when challenges are going to rise ahead of time. This results in less direct labor requirements, higher customer fill rates, and significantly more operational efficiency.

Keith Moore is the Chief Product Officer at AutoScheduler.AI. Before venturing into the supply chain business, Moore was a Director of Product Management at SparkCognition. He helped raise over $120 million in capital and grow the business to one of the most prominent start-ups in the Austin, Texas area. Moore attended the University of Tennessee, where he received a B.S. in Mechanical Engineering. He was terrible at mechanical engineering and proceeded immediately to a life in software.