Automation of forecasting with TimeGPT saves the team approximately 1 hour per day, or roughly 30 hours per month
Accurate refill predictions ensure that dispensers remain operational throughout business hours
Re:Drink is a Munich-based startup offering a bottleless smart beverage dispenser for offices, universities, and business spaces. The countertop IoT unit dispenses filtered tap water, enriched with natural flavor concentrates, and lets users customize each drink's flavor intensity, sweetness, and carbonation. More details can be found on their website, Re:Drink — Die Flaschenlose Getränkelösung
Aligned with a strong sustainability mission, Re:Drink eliminates more than 51,000 bottles per year and cuts CO₂ emissions by 93% at client sites through its packaging‐free approach. Early deployments have seen hundreds of people using the dispenser daily, as highlighted in Re:Drink's usage data
The team's vision of decentralizing bottling plants aims to reduce beverage supply logistics by approximately 85%, providing a 24/7 refreshment station that enhances workplace experience and supports sustainability goals
Re:Drink integrates IoT hardware with cloud intelligence to enable real-time monitoring and 'flavor forecasting', predicting when flavor concentrates and CO₂ cartridges will deplete
Re:Drink needed to accurately predict when its smart dispensers would require refills of flavor concentrates and CO₂ cartridges to ensure 100% uptime without overstocking
Re:Drink implemented Nixtla's TimeGPT to replace its manual forecasting models, achieving rapid, accurate, on-demand predictions through a simple API integration