Unilever Ice Cream improves forecast accuracy 10% with AI weather-driven demand planning

Ice cream is one of food manufacturing's most weather-sensitive categories: demand can shift dramatically within hours based on temperature, making annual production planning exceptionally difficult. Unilever Ice Cream operates across 60 countries with 35 factory production lines and an estimated 3 million freezer cabinets — a network where even a 1°C temperature variation in a European summer month like June can materially alter sales forecasts. Producing against static annual plans in such a volatile environment created persistent misalignment: excess inventory during unexpectedly cool periods and stockouts during sudden heatwaves. The resulting waste, elevated production costs, and degraded service levels made a structural forecasting overhaul unavoidable.

Unilever deployed machine learning and predictive analytics to integrate weather data alongside a broader set of demand signals — including promotional calendars, historical consumption patterns, and regional purchasing behavior — into a unified forecasting model. The system generates probability-weighted volume forecasts across multiple temperature scenarios, giving long-term planning teams a range of outcomes rather than a single point estimate. On the short-term side, an inventory reallocation capability allows supply chain planners to redirect stock in near-real time toward markets experiencing unexpected demand spikes such as heatwaves. In parallel, AI-enabled smart freezers — now numbering 100,000 units — feed continuous point-of-sale and replenishment data back into the planning system, closing the feedback loop between the factory floor and the retail shelf. Together, these capabilities give planners visibility spanning 12-month production schedules down to individual cabinet-level replenishment decisions.

In Sweden, the primary proof-of-concept market, forecast accuracy improved by 10%, enabling production lines to be adjusted closer to actual expected demand and reducing unnecessary output costs. Across the broader network, service levels reached what the company describes as world-class, with Unilever ranking in the top tier in the majority of its markets. The AI-enabled smart freezer fleet delivered a distinct commercial uplift: the 100,000 connected units increased retail orders and sales by up to 30% through more precise replenishment decisions. Externally, the cumulative supply chain performance was validated when Gartner named Unilever one of four Supply Chain Masters in 2024 — a recognition the company has held for six consecutive years.

• Weather probability modeling is a high-ROI forecasting input for seasonal food categories — even small temperature variations can outweigh conventional demand signals like promotions or pricing.
• Combining long-horizon scenario planning with a short-term stock reallocation capability addresses both structural and reactive demand challenges within a single integrated system.
• Connected retail assets (smart freezers, cabinets) create feedback loops that improve both inventory efficiency and top-line sales simultaneously, making them a dual-value investment case.
• Piloting in one market before global rollout generates a measurable proof point — such as Sweden's 10% accuracy gain — that builds organizational confidence for broader deployment.