Unilever

In consumer packaged goods, even modest gaps in on-shelf availability translate directly to lost sales and strained retailer relationships. For Unilever — a global CPG manufacturer with a broad product portfolio spanning food, beverage, and personal care — maintaining reliable replenishment at Walmart Mexico, one of its largest retail customers, posed a structural challenge. Traditional supply chain processes created data silos between manufacturer and retailer, making it difficult to align forecasts, inventory positions, and store execution in near-real time. Without visibility at the day-and-store level, replenishment signals were slow and imprecise. The result was OSA rates that fell short of what a high-volume retail partnership demands, leaving both revenue and category leadership on the table.

Unilever addressed this by building a proprietary machine learning and predictive analytics platform internally branded as 'One Supply Chain.' The model integrates point-of-sale data, inventory management feeds, and demand forecasts from both Unilever and Walmart Mexico into a single, synchronized data layer — eliminating the information lag that historically delayed replenishment decisions. By generating forecasts at day-and-store granularity, the system produces precise replenishment signals rather than the aggregated averages that obscure local demand patterns. Promotional planning and execution are also synchronized within the model, aligning upstream production with in-store events. Unilever first piloted the solution in its nutrition category in 2022, validating performance before expanding the rollout across all in-store Walmart Mexico products — a staged approach that contained integration risk at each phase.

The Walmart Mexico deployment achieved on-shelf availability rates exceeding 98%, the program's defining outcome. Within less than a year of launch, the nutrition category recorded strong category growth — realized simultaneously with a reduction in inventory levels, demonstrating that higher service levels do not require carrying more stock. Additional reported improvements include:

• In-stock and fill rates increased across the full product range
• Forecast accuracy improved through real-time POS synchronization
• The Unilever–Walmart Mexico relationship shifted from a traditional vendor model to a fully collaborative, data-sharing supply chain partnership

Based on these results, Unilever is now extending the model to key global retail customers, smaller regional retailers, and digital commerce channels.

• Real-time, bidirectional data sharing between CPG manufacturer and retailer is the foundational prerequisite — AI cannot compensate for upstream data gaps or siloed systems.
• Day/store granularity is the minimum resolution needed to generate actionable replenishment signals; aggregated forecasts obscure local demand patterns.
• A single-category pilot limits integration risk and builds internal confidence before committing to a full product-range rollout.
• Synchronizing promotional planning within the supply chain model is critical — unplanned demand spikes from promotions are a leading cause of preventable stockouts.
• High OSA and lower inventory are not inherent trade-offs; a well-calibrated AI model can deliver both simultaneously.