Hoerbiger uses Machine Learning




Advanced Analytics for IoT



Reading time

3 minutes

Hoerbiger assigns materials automatically and correctly with the help of machine learning.

How A1 Digital approached the solution


Data sample as a base

Based on a data sample with 100 percent correctly assigned materials, A1 Digital created a model to teach the system to distinguish between correct and incorrect assignments.


Learning algorithm created

A learning algorithm is generated, which – applied to known and already classified cases (the database) – calculates structures.


Enabled evaluating algorithm

These newly learned structures enable another, evaluating algorithm to assign a new and previously unknown case to one of the known categories based on the observed attributes and their characteristics.

Hoerbiger case study robert fruhwirth

"Depending on the number of products, the classification of products into appropriate categories is associated with extensive manual work without machine learning techniques. We have seen that machine learning is generally able to select the right product group based on detailed material descriptions."

Robert Fruhwirth Head of Purchasing Process and Spend Management

The Customer:

The trading company Hoerbiger & Co, founded by Hanns Hörbiger in Vienna in 1925, has developed over the years into Hoerbiger Holding AG, a group based in Zug, Switzerland, and represented in over 50 countries with 140 production and service locations. Today, Hoerbiger holds a leading position worldwide with performance-defining components in the fields of compression technology, drive technology and hydraulics. The history of the company is characterized by a technical pioneering spirit that has set standards with innovations. Even when it comes to digital progress and the use of artificial intelligence to optimize processes, the company has retained this pioneering spirit to this day.

The Challenge:

All materials and components - from safety shoes to screws - that are purchased in the group are assigned to so-called categories (product groups) according to certain criteria. A product group tree is stored in the Hoerbiger Group's ERP system for this purpose. The individual materials are assigned manually or according to certain logics that assign the materials to the product groups. Over time, the system has turned out to be a weak point, as the hit rate for the assigned product groups is unsatisfactory. The intuitive solution of manually checking all products for the correct categorization was discarded due to the large number of products.

The Solution:

Triggered by a lecture on the subject of machine learning (ML), those responsible at Hoerbiger came up with the idea of ​​using artificial intelligence to create a system that automatically and correctly assigns the materials to the product groups based on the description. However, the optimal application of ML requires appropriate know-how and suitable tools. Hörbiger therefore commissioned A1 Digital International GmbH to develop and implement a machine learning solution as part of a proof of concept.

Similar Case Studies