Machine Learning for Smart Purchasing: Vetter Pharma

Industry: Pharmaceutical industry

Products: Aseptic filling and packaging of prefilled injection systems such as syringes, cartridges, and vials

Employees: approx. 6,000 (2023)

Headquarters: Ravensburg, Germany

Website: www.vetter-pharma.com

Highest possible quality – patients first

Vetter, a global company, traces its roots back 60 years, when it was founded as a pharmacy. The manufacturer now specializes in aseptic filling and packaging of syringes and other injection systems. From the initial development phase of new medicines all the way to their supply to markets, Vetter provides support to the manufacturers of drugs for the treatment of various conditions. Some 6,000 highly skilled employees and the use of intelligent robots in aseptic production ensure adherence to the highest quality standards. Focusing on innovation and automation, Vetter also relies on state-of-the-art technologies in its purchasing activities.

“First-time-right” – greater efficiency in the purchasing process

There are two objectives that form part of ongoing system optimization. The first is to increase the “first-time-right” rate in the recording of purchase requisitions. The second is to enable higher transparency of purchasing volumes and in so doing, support strategic decision-making. To this end, the product group concept has undergone comprehensive adaption to the new requirements of the market environment. The number of product groups has also increased considerably as a result. At the same time, it is important that users get technical support so that they can reduce manual errors and minimize the effort required to create purchase requisitions. To achieve these goals, a Machine Learning service is now being used, which supports users in selecting the product group and offers suggestions.

Machine Learning – flexibility in times of rapid change

In addition to the revision of the product group concept by the specialist department, a procedure model in the SAP Data Attribute Recommendation (DAR) service was also taught in using training data records. The model provides a prediction and a probability value based on the individual data input. Users can view this via a button that displays three product group suggestions via a popup, along with the probability value based on the AI model. This allows users to choose a suggestion and continue with the process. This model is automatically taught in at regular intervals with further training data so that it can adapt to a dynamic business environment following a defined teach-in phase. For the popup to work, an interface to the SAP DAR service was defined via the SAP BTP and integrated in the purchase requisition transaction.

Potential – additional usage areas identified

Despite the extensive expansion of the product groups for the modern market environment, it has proved possible to further improve the “first-time-right” rate for purchase requisitions. The DAR Service yielded considerable positive feedback from requesters. On account of the success, further usage areas are now under consideration, such as to support users with maintenance orders using Machine Learning. There is nothing standing in the way of a digital, intelligent future.