JapanFood and restaurants2025 – ongoingUmios (formerly Maruha Nichiro)
Sales-plan forecasting with the Chronos-2 time-series model
Umios, formerly Maruha Nichiro, reported a 95% match between Chronos-2 forecasts and its sales plans before going into production
Problem
Sales planning for the company's business-use products took many staff hours each year, and excess inventory also needed manual handling. The company set a target of cutting 4,200 hours a year.
Approach
The company used Amazon's Chronos-2 time-series model through an API on AWS. A workflow pulls shipment history and inventory data, and forecasts go to a BI tool where planners review them before they enter the planning system. The model was received in December 2025, tested from February 2026 and put into production in June 2026. Seasonal items showed larger errors, so the company set a safety-stock approach for them.
Results
Takeaway
The company agreed an accuracy threshold before the test and kept planners as the final decision makers. The 4,200-hour saving was still a target.
Categories
- Size
- Large enterprise
- Industry
- Food and restaurants
- Business function
- Logistics and supply chainProduction and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Company's own claim
Sources
Source summaries
The article covered a company talk at AWS Summit Japan 2026. It reported the use of Chronos-2 for sales-plan forecasting, a 95 percent match rate in verification and a June 2026 production start. It gave a 4,200-hour annual reduction as a goal.
Full details are in the original sources.