Skip to content
Data Science & AI

Haubis and Wewalka — Efficient Demand Planning with AI-Driven Forecasting

In a nutshell

We partnered up with Haubis and Wewalka to implement AI-powered demand forecasting models that harness historical data, expert insights, and customized variables to optimize their inventory management, minimize waste, and foster data-driven business decisions and plans.

Haubis & Wewalka
Business Challenge

Inaccurate demand forecasting resulting in inefficient inventory management, i.e. under- or overproduction, stockpiling, and financial losses.

Project overview


06/2023 – ongoing

Team setup

Python (Darts, XGBoost, scikit-learn), Azure, GitHub


Python, HuggingFace, Azure, GitHub


AI & Data Science

About our clients


Haubis GmbH is a leading Austrian bakery specializing in high-quality breads, pastries, and other baked goods. Fusing traditional baking techniques with modern trends, Haubis is known for its commitment to delivering premium products and using high quality ingredients. With a rich history dating back to 1902, Haubis’ products are nowadays sold in numerous countries worldwide, and the company has a strong reputation for quality and innovation.


Wewalka is a leading Austrian manufacturer of fresh doughs known for their quality and freshness. Since 1987 and offering a single product only – a puff pastry rolled on parchment paper – the company’s product range has grown rapidly with various products (incl. brand Tante Fanny) now being sold in 30 European countries, making Wewalka one of the major players on the fresh-dough market.

Challenges & objectives

Both Haubis and Wewalka faced significant challenges in accurately predicting demand for their products. Their forecasting methods were solely based on internal expert knowledge, which was not always sufficient to account for market trends, seasonal fluctuations, and other external factors. This led to several issues, including overproduction and excess inventory, which also affected companies’ storage capacities and transportation costs, and in the end, led to financial losses.

To address these challenges, Parkside Interactive embarked on a collaborative endeavor to develop and implement AI-driven forecasting models for Haubis and Wewalka. These models would harness historical data, expert insights, and customized variables to capture the intricacies of each company’s business and market dynamics. By improving demand forecasting accuracy, we aimed to achieve the following objectives:


Reduction of overproduction

More accurate forecasts would allow Haubis and Wewalka to produce the right amount of products to meet demand, eliminating the need for excess inventory.

Minimization of inventory levels

With more precise demand predictions, the companies could maintain optimal inventory levels, reducing storage costs and transportation expenses.

Enhancement of profitability

By reducing waste and optimizing inventory management, Haubis and Wewalka would improve their financial performance and allocate resources more effectively.

Our solution

Our approach encompassed the entire data science lifecycle, from data acquisition and preprocessing to model training and deployment. We worked closely with the Haubis and Wewalka teams to extract relevant data from their internal systems, clean and prepare it for analysis, and develop AI models that effectively captured demand patterns. To make sure that our models are tailored to each company’s specific business needs, we organized workshops with experts from both companies to identify key variables that could significantly influence demand and incorporated those into our ML models afterward.

End results & future plans

Our collaboration with Haubis and Wewalka resulted in the successful implementation of AI-powered demand forecasting solutions for both companies. The ML models we created provide automated forecasts for key business users on a daily basis, enabling them to make informed decisions about production planning, inventory management, and supply chain optimization. The implementation of these models has led to significant improvements in demand accuracy, reduced inventory levels, and minimized waste.

In the future, the accuracy of the model can be further improved by providing more edge cases of different classes. We have also developed a concept for deploying our solution via Azure, which enables seamless integration into Rapp AG’s existing infrastructure. This end-to-end approach facilitates troubleshooting and maintenance in the continuous use of this machine learning application.