United StatesFood and restaurants2019 – ongoingStarbucks
Deep Brew personalization and store operations AI
Starbucks used one AI platform for app recommendations, staff scheduling and inventory, but has published no results
Problem
Starbucks serves more than 100 million customers a week across about 30,000 stores. Offering each customer relevant suggestions, and setting barista schedules and inventory store by store, was hard to manage centrally.
Approach
By 2019 Starbucks ran a reinforcement-learning platform, hosted on Microsoft Azure, that suggested orders in its app based on past orders, local inventory, weather and time of day. It tested drive-through menu boards that change their suggestions using store transaction data. Under the name Deep Brew, the company also used its predictive models to support barista scheduling and automate daily inventory orders.
Results
Takeaway
The platform links marketing and store operations through shared data. Starbucks has not published the sales or labor effect of Deep Brew.
Categories
- Size
- Large enterprise
- Industry
- Food and restaurants
- Business function
- Marketing and salesLogistics and supply chain
- Innovation type
- Product and service innovationProcess innovation
- Adoption
- Built in-house
- Evidence
- Vendor's case study
Sources
Source summaries
Microsoft reports that a reinforcement-learning platform on Azure gives Starbucks' 16 million active Rewards members order suggestions based on local inventory, popular items, weather, time of day and past orders. It reports a drive-through menu board using store transactions and more than 400 store-level criteria, then in testing, and quotes Starbucks executives on using machine learning for inventory and barista schedules.
Nation's Restaurant News reports that Starbucks said in its fiscal second-quarter 2021 earnings call that Deep Brew calculates personalized offers and suggestions across hundreds of US stores, automates daily inventory orders and supports staff scheduling. CEO Kevin Johnson said AI provided the underlying predictive models; Starbucks gave no figures for the AI's effect.
Full details are in the original sources.