Toridoll Holdings has deployed Fujitsu’s AI demand forecast service at 823 noodle shops of Marugame Udon across Japan, to enable accurate forecasting of the number of customers and sales by day and time for each shop based on weather data and point-of-sale (POS) data.
Fujitsu will help digitally transform operations for shop managers, optimising complex manual processes including the ordering of products and food quantity planning to deliver greater overall efficiency, help reduce food loss, and realise efficient energy management through the optimisation of staff allocation and air conditioning in shops.
Moving forward, Fujitsu and Toridoll will continue to leverage AI and other leading-edge technologies, as well as various data and business know-how, to promote the digital transformation (DX) of Toridoll’s’ restaurants globally.
In 2021, Toridoll started testing the effectiveness of a new AI demand forecast service system for predicting the customer numbers and sales of its Marugame Udon brand by shop, day, and time.
This was done to realise two goals of its vision — automated shop management using AI demand forecasts, and energy management system using IoT. Toridoll decided to deploy the new solution at all of its 823 shops in Japan.
The AI forecast service is based on Fujitsu’s AI demand prediction solution “Fujitsu Business Application Operational Data Management & Analytics Demand Forecasting SaaS,” which enables users to predict future customer and sales numbers with high accuracy based on various data held by companies, including POS data, sales calendars, and sales promotion campaigns, as well as weather data.
Toridoll and Fujitsu will use the AI demand prediction service to automate and improve work schedules and optimise order processes, food quantity planning and energy use.
Fujitsu’s solution promises stable and accurate demand forecasting using AI and machine learning technologies.
Using a model of Fujitsu Laboratories for dynamic ensemble forecasting that leverages AI and machine learning technology to imitate human thought processes to make predictions from data characteristics, Fujitsu realised an optimal combination of multiple demand prediction models through automatic tuning.
In this way, the service can offer both stable and highly accurate forecasts without the need to select from different forecasting methods by leveraging a learning model that accurately captures the characteristics of individual prediction objects that change according to various factors including periodicities, external factors and trends.
The solution also promises easier cooperation to make use of forecast data in various operations.
As Fujitsu offers the forecast service via the cloud, required forecast data can be easily linked with various SaaS applications and APIs running in the cloud, thus supporting the use of forecast data in various planning operations including order placement, production planning, and work scheduling.
