Walmart

Walmart operates one of the world's most complex retail supply chains, spanning thousands of stores, distribution centres, and fulfilment centres serving hundreds of millions of US households. Historically, inventory management was conducted store-by-store, creating fragmented visibility, duplicated safety stock, and limited ability to respond dynamically to demand shifts or sourcing disruptions. As sustained tariff pressure intensified and e-commerce expectations accelerated, this decentralised model became untenable. Walmart needed to extend sub-three-hour delivery guarantees to 95% of US households while absorbing roughly 500 million marketplace SKUs — all without proportional increases in owned inventory or unit fulfilment costs, and against a backdrop of rising labour and input expenses.

Walmart deployed machine learning and predictive analytics across a unified inventory orchestration spine, integrating real-time store-level data with automated DC and FC feeds. Handheld devices and computer vision were rolled out to more than one million associates, providing live stock mapping that feeds centralised routing and sourcing decisions. Automated regional DCs now deliver to approximately 60% of US stores on a predictable, palletised basis, while more than half of e-commerce FC throughput is handled by automation. ML models steer replenishment and sourcing centrally, treating owned inventory, marketplace SKUs, and node capacity as a single pool. Individual stores are assigned explicit roles — fulfilment-heavy nodes for high-density digital pick operations, or experience-priority locations with lighter brown-box volumes — with AI determining which node serves which demand based on speed promise and delivered cost.

The orchestration programme delivered measurable gains across inventory efficiency, cost, and delivery reach. US owned inventory rose only 2.6% against mid-single-digit sales growth, despite absorbing approximately 500 million marketplace SKUs. Globally, inventory growth was held to roughly half the rate of sales. Automated FCs cut handling time and unit shipping costs by double-digit percentages while reducing fresh-category waste. Walmart can now reach 95% of US households within three hours, with sub-one-hour coverage across many catchments:

• Inventory growth vs. sales growth: ~2.6% US owned inventory vs. mid-single-digit sales
• FC cost reduction: double-digit % in handling and unit shipping
• Delivery reach: 95% of US households within three hours
• Price rollbacks increased 20%+ to over 7,000 items, more than half in grocery

• Assigning explicit node roles — fulfilment-heavy vs. experience-priority — to individual stores is a prerequisite for AI-led routing to meet speed promises without duplicating safety stock network-wide.
• Marketplace scale (~500M SKUs) can extend assortment breadth without proportional owned inventory growth only when ML actively curates range and routes fulfilment to the correct node.
• Automation in FCs creates double-digit cost reductions that simultaneously fund subsidised pricing and rapid delivery, making speed and value mutually reinforcing rather than competing objectives.
• Real-time associate-level visibility via handhelds and computer vision is the foundational data layer; centralised AI routing decisions are only as reliable as the store-level signal feeding them.