FEV Makes Software Quality Measurable

Metrics-based Strategies for Quality Assurance of Automotive Embedded Software

6. August 2017 | Software & Testing Solutions

Digitalization and high innovative demand drive future key technologies in the automotive industry. Increasing software complexity is met by rigorous quality and safety standards, but strict resource constraints on projects are limiting time spent on quality assurance (QA). Therefore, FEV has developed a tailored, optimized QA strategy that is scientifically sound to improve efficiency and quality in software projects. Based on data derived from 13 customer projects, the impact of strategic decisions on the quality of the resulting software was examined, thus leading to a better understanding of how high quality can be achieved and quantified appropriately.


Focus of the Study

Two major subjects are the primary focus of FEV’s research. First, the aim is to analyze, formalize and then capture the quality of a software product using the appropriate metrics. The second goal is to identify how specific properties of the quality assurance strategy influence the outcome of a project.

A pragmatic approach to defining software product quality is its segmentation into intrinsic and extrinsic facets. Extrinsic quality is most often judged on the basis of the customer’s satisfaction, whereas intrinsic quality can best be quantified using established product quality models, such as that found in the ISO 25010.

Key of the Study

For the purpose of the initial study, FEV limited the scope of investigation to certain properties proposed by the ISO 25010. Relevant quality characteristics were then quantified using a set of metrics. Most notably, the requirements conformity and residual defect ratio were considered.

The data from 13 automotive software projects was gathered and analyzed in order to validate the correlation between metric and de facto product quality and to identify any relationships to specific QA strategies. Statistical models, such as correlation coefficients and ANOVA (Analysis of Variance), were considered for analysis purposes.


The research showed that customer satisfaction is strongly correlated with both requirements conformance (positive correlation) and residual defect ratio (negative correlation), which indicates a strong linear relationship between these metrics.
We make the reasonable assumption that intrinsic and extrinsic product quality strongly correlate with each other. Then, both metrics – requirement conformance and residual defect ratio – appear to be suitable indicators for intrinsic product quality, since customer satisfaction is regarded as the major marker for extrinsic quality in literature.


Lastly, the impact of different quality assurance strategies on a project’s outcome was analyzed by first surveying which test methods were applied (respectively), and then examining the relationship between these properties and the identified quality indicators. First results here are promising, and FEV is confident that it will enable the profound derivation of an effective and efficient QA strategy in upcoming projects.

Graphic - automotive software

Relevant software products characteristics and metrics evaluated.