Canteen goes AI - with Delicious Data

More than 2,600 kilos of food saved at three locations within six months: this is the result of the pilot project run by ERGO Gourmet, the company that operates the ERGO canteens, and the start-up Delicious Data. Now the AI is being rolled out at all ERGO locations in Germany.

How does today's weather forecast affect the number of curry sausages ordered in Berlin? Do the people of Munich eat more salad before the holiday? How do the people of Cologne feel about bibimbap on the first day of the school holidays? The artificial intelligence (AI) of the start-up Delicious Data can provide reliable answers to these and other questions in real time - and the kitchens immediately adjust the quantities produced. It demonstrated this in a pilot project that ran with such great success between July and December at the ERGO locations in Berlin, Cologne and Munich that the AI is now being introduced as standard at all locations.

"With Delicious Data, we can act sustainably and manage in an environmentally conscious way," says Barret Jähn from ERGO Gourmet. With this idea to avoid food waste, the 2020 team won an internal competition and received the funding for the pilot project.

Forecasts in real time

Delicious Data was "fed" with the menu plans and sales figures of the Berlin, Cologne and Munich locations from previous years. Based on this and taking into account external factors such as weather, holidays or public holidays, Delicious Data can predict very accurately how many guests will choose which dishes. Through constant feedback from the kitchen, the AI updates its forecast in real time about how many more sausages should be put on the grill and whether the tomato salad in the counter is sufficient. Even if the Kaiserschmarrn is spontaneously replaced by a veggie burger with fries one day, Delicious Data predicts how many burgers will be consumed and how this change will affect the other dishes.

Together with the start-up, ERGO Gourmet has also developed a new food waste monitoring system. The food waste at each location is measured and immediately analysed. Even broken down to each plate. "This enabled us to reduce food waste by 27 percent in the pilot phase," Jähn reports. But Delicious Data not only saves food, but also CO2 and water, as a graph shows (German only):

Delicious Data is just one of several sustainable and innovative projects by ERGO Gourmet. The canteens have already joined the "toogoodtogo" app since 2019 and have been testing the digital reusable solution "Vytal" to avoid disposable packaging since February 2021.

Text: Stefanie Neumann

What specifically makes up your business model?

Belser: We founded Delicious Data with the aim of using advances in AI to increase sustainability in out-of-home catering.

Breuninger: To do this, we have developed a deep-learning algorithm that calculates a precise forecast for future sales from the history of customer data, planned meals and other external factors - such as weather, holiday periods or public holidays. This makes it possible to buy according to demand and plan production well accordingly. There is also the option of accurately recording food waste and visualising this data together in a dashboard. This creates additional transparency and makes it possible to optimise processes in concrete terms. Our system is now used at more than 250 locations. 

Explain: How can AI help with the issue of food waste?

Breuninger: Our AI-based sales forecasts map very precisely and realistically which dishes the customers would like to have, so that the purchase and then the preparation of the dishes can already be optimised here. A dynamic calculation based on the grown database leads to very accurate forecasts that are also updated during the course of the day. In this way, overproduction can be minimised even in large canteens at the end of the value chain.

What exactly does Food Waste Tracking mean in this context?

Belser: In addition to forecasts, our tool also offers food waste monitoring. This enables users to precisely record food waste and to visualise and evaluate its development via a dashboard. This reveals further savings potential. In concrete terms, the size of the respective portions or the choice of assortment can then be considered during planning.

Breuninger: Practically, this feature is based on a web app in which the weighed waste is directly recorded and displayed in our analysis dashboard. So there is no need for any additional hardware or other IT integration from external systems. This feature is the result of our successful cooperation with ERGO Gourmet.

According to a study by the German Federal Ministry of Food and Agriculture, 20 per cent of food in canteens and cafeterias ends up in the rubbish. How much of this could be avoided with AI support?

Belser: Quite a bit. The savings potential is made up of several factors that we can influence with our software. First of all, there is the planning accuracy, which we can optimise through the forecasts and which reduce food waste already in purchasing. Our experience shows that the planning accuracy is 40 per cent better. Secondly, the survey of food waste contributes to the fact that we have been able to reduce food waste at our customers by 30 per cent with further analyses.

What is the cooperation with ERGO Gourmet's company catering and what have you been able to achieve there so far?

Breuninger: At this point it is really interesting to look at the concrete figures that we have already been able to optimise for ERGO Gourmet: Because in the 6-month pilot phase alone, almost 7,000 meals could be saved. This corresponds to a climate impact of 8,651 kilograms less CO2 and 380,650 litres less fresh water wasted. We are pleased about this commitment to more sustainability and would like to take this opportunity to thank ERGO for the very good, cooperative partnership.

Which industries can you support with your AI concept?

Belser: We want to make a positive impact with our solution in all areas of the food value chain and support companies on the path to digital transformation. Wherever demand is uncertain and fluctuations cannot be optimally mapped and incorporated manually, AI offers opportunities to save resources. In this way, we can help not only canteens and cafeterias, but also bakeries, snack bars and system gastronomy.

What developments are in store for the future?

Breuninger: We are currently working on the intraday forecast, i.e. a forecast for the current day. This involves calculating over the course of the day how many goods will be needed and whether additional production will be necessary on site. This evaluation can be controlled with a simple "to-do" app, so that here, too, work processes are simplified and production takes place according to the calculated demand, which optimises the use of food.

Interview: Benjamin Esche