Restaurants face a big problem, that they don't know how many guests are coming the next day or next week. This leads to different problems like:

• Foodwaste
• high personnel costs
• unhappy customers

Our team tried to solve this problem by forecasting the amount of people for noon and evening for a restaurant located in Zürich. This forecasts are afterwards displayed in an App for the kichen chef and the chef de service.

We took data from a reservation system as our brain how the demand was in the past. Together with influencing factors like holidays, weekdays, seasonality we calculated our forecasts. The output of those forecasts is being written in the API of Prognolite. The App, that we developed with an experienced designer connects to the Prognolite API calls the forecasts, weather icons and holidays and visualizes it. We found out that the predictions for the evening are much better than for noon and found the reason in employees who don't register all the walk-ins.

In addition we developed an ExcellPrediction prototype. The Prototype uses turnover data from the past year(s) to make a forecast for future sales, with respect to any relevant parameters such as weather or holidays. Based on the fact that turnover (daily revenue) is correlated (0.95) with table reservations and number of dishes sold, it can be used to make a prediction on how much food restaurant owner needs to buy for the next day, week or month. The model also uses rolling window to adapt the forecast to any new unexpected circumstances.

Team members:

• Roland Brand
• Claudia Fricker
• Eugene Orlov
• Olga Matveeva
• Ewa Guminska
• Etienne Soguel-dit-Piquard
• Roman Lickel
• Simon Michel

Our App:

Presentation: