More product quality with Quasar

Software-intensive products rarely fail because of individual lines of code. More often, it is conceptual weaknesses that only become apparent late in the process—for example, in early requirements, ad-hoc architectural decisions, or a lack of documentation. Correcting these issues later is both time-consuming and costly. These conceptual weaknesses have a major impact on the quality of the final product, especially when longevity is considered. Addressing these weaknesses conflicts with the development of new features. Furthermore, the extent of the problem is often unknown—not because of a lack of will, but because systematically and continuously measuring many quality aspects simultaneously alongside day-to-day operations is hardly feasible. This is exactly where Quasar comes in.

Quasar is an approach developed by Fraunhofer IESE for comprehensive, continuous quality measurement throughout the entire software lifecycle. Parts of it are already in productive use, such as for assessing the quality of digital solutions on the marketplace Deutschland.Digital. The ultimate goal, however, is to enable companies to implement such an approach as their own internally operated tool.

The core of Quasar is not a single, generic language model. Established quality models based on the Goal-Question-Metric (GQM) approach precisely define what should be measured in each artifact. Specialized modules each address a specific quality aspect, using generative AI in a targeted manner for individual, clearly defined measurement points. The resulting recommendations are methodologically transparent and technically justifiable. This is the prerequisite for quality assessments to actually guide decisions in practice.

Quasar doesn't just look at code. Its quality assessment begins with early-stage artifacts: requirements, concepts, designs, architecture documents, and other documentation are considered, as are repositories and finished software products. The earlier a quality issue is identified, the less effort it takes to fix it.

Quasar was designed from the outset to process internal company data exclusively in a secure manner. The underlying language models can be run locally or in trusted, hosted data centers. There is no reliance on external cloud services.