2023 Edelman Award Winner Reprise: Walmart uses O.R. to enhance retail truck routing and load planning

The 2023 Franz Edelman Award was awarded to Walmart during the Edelman Gala at the 2023 INFORMS Business Analytics Conference in Aurora, Colorado, for its use of operations research (O.R.) to enhance retail truck routing and load planning. 

It would be difficult to find someone who has never heard of Walmart, but few know the behind-the-scenes of its growing supply chain network. The first Walmart store opened in 1962 in Rogers, Arkansas, and only a few years later increased to two dozen stores. The first Walmart Supercenter opened in the 1980s, which became nationwide in the 1990s. From the 2000s to today, Walmart operates more than 10,500 stores and clubs in 20 countries and e-commerce websites, which as you can imagine, carries a massive supply chain network.

Walmart offers an assortment of commodities online, in store and through a mobile app. Most recently. Walmart has scaled its in-store pickup and in-home delivery across the U.S. The goal was to transform groceries while keeping their signature everyday low prices, as well as to keep improving processes and the supply chain. Reaching these goals had four stages:

1. Network design – What would be the end-state network in 10 years?
2. Network transformation – Step-by-step approach
3. Route & load planning – From distribution centers (DCs) to stores
4. Simulation – Leveraging models and algorithms for strategic decision-making

In the end, Walmart’s goal was to make a positive impact on business and the environment while supporting its associates and customers.

The answer was supply chain analytics for both e-commerce and brick-and-mortar models – an omnichannel approach to meet customers where they’re at digitally. In-store shoppers want a digital presence with the ability to order online and pickup in-store. The stores and DCs needed operations research (O.R.) to bring this to life and to the customer at both the network and distribution level.

Walmart embarked on journey to improve supply chains to service this omnichannel customer, which required flexibility and the need for multifunctional assets to support multiple shopping occasions. Introducing automation in all assets would help improve productivity and reshape what an associate does inside DCs and fulfillment centers (FCs). High-level O.R. and decision sciences were used.

Prakhar Mehrotra, VP of Applied AI at Walmart, said, “O.R. is part of everything we do at Walmart. It’s unique because O.R. was used on the strategy and execution side, and we brought them together.” O.R. helps with longtime, irreversible decisions that a company has to make, such as where to open a warehouse, as well as daily decisions to help supply chains, such as routing trucks, etc. Decisions made on strategy side affect what happens on the operations and execution side. The data science team at Walmart used O.R. on both sides. In particular, the three key areas: network strategy; execution: route and load planning; and simulation platform.

4 step hierarchical process for end-to-end supply chain:

1. Outlining strategic design network: Apply advanced analytics and O.R. to a nontraditional area: strategy. Goals were to design a future network and create a transformation roadmap (like when to build new or close existing DCs). Walmart built two models (end-state and transformation) and ran hundreds of scenarios with these models, allowing leaders to evaluate based on macroeconomics to make decisions. Multiple-scenario evaluations allowed for accelerated decision-making. The team created dashboards for KPIs to get an idea of how the scenarios work. And ultimately, the work done for this part of project was transformable to other areas.
2. Plan network transformation to get from current to future state (perhaps a few years in the future).
3. Plan resources and capacity (store-to-DC alignment and capacity planning at each node).
4. Plan routes & loads – pick, load, drive, unload

Jing Huang, director of applied science, transportation and last-mile delivery at Walmart, discussed the execution of load planning for grocery outbound distribution. Once pallets are stocked, the team gets started on a routing plan. Each route generated at this stage is reviewed by the transportation command center (TCC), who revises routes for feasibility. The best load plans are generated for best operations and reviewed by the TCC. At that point, pick-and-load activity begins using the plan made, which can be adjusted on the fly. Drivers drive trucks of pallets along routes and plans. The Load Planner generates the most efficient route to arrive in the time window. Perishable products require compliance throughout storing, moving etc. Therefore, the Load Planner must solve for the routing problem with consideration of loading problem and constraints including safety, compliance and more.

Next up: Simulation for store-to-DC alignment, trailer configuration and deliver frequencies. Walmart built a simulation platform allowing users to quickly review scenarios and make optimal decisions. Fast-running times allow for better network design and the Load Planner to make more optimal decisions.

Adoption and User Feedback

Kim Crozier, director of data science at Walmart, discussed merging e-commerce with in-store using next-generation automation capabilities. Mixed-integer modeling and programs were used to help establish the optimal future state for Walmart. After four years, they were able to present the transformation to better work environments for DCs, and associates and customers in stores. At that time, the team had also successfully tested and piloted distribution planners and built a successful Load Planner. It is highly configurable and allows for quick changes after assessment.

“Walmart is huge and it’s easy to work in silos, but the Load Planner is a tool that will inherently help us be better – from picking up a crate to delivering to a customer’s fridge,” said Megan Ferguson, VP E2E Delivery Acceleration at Walmart. It was built in a way to integrate with other Walmart products and technologies, resulting in improved level of service to stores.

All of these products coming together is the competitive angle for Walmart and will optimize processes to help them win long term. “We are a people-led, tech-empowered company and Load Planner is the epitome of what that looks like,” said Ferguson.

Breakthroughs and Impact

The Edelman-winning project was the collective effort of a large number of Walmart associates in the past four years. It took a diverse group of people to come up with a unified solution that proposed decision-making across the supply chain using highly scalable and flexible network strategy models. The team simultaneously solved routing and loading problems with rich operational constraints, as well as solved the variable-sized multicompartment container loading problem in a millisecond.

In fiscal year 2023, with the full network rollout of Load Planner, Walmart saved $91.5 million and prevented 98.6 million pounds of CO2 emissions by avoiding 33 million miles and 108,000 truck routes.