Abstract : We present Cardiogram, a visual analytics system that supports automotive engineers in debugging masses of traces each consisting of millions of recorded messages from in-car communication networks. With their increasing complexity, ensuring these safety-critical networks to be error-free has become a major task and challenge for automotive engineers. To overcome shortcomings of current analysis tools, Cardiogram combines visualization techniques with a data preprocessing approach to automatically reduce complexity based on engineers' domain knowledge. In this paper, we provide the findings from an exploratory, three-year field study within a large automotive company, studying current practices of engineers, the challenges they meet and the characteristics for integrating novel visual analytics tools into their work practices. We then introduce Cardiogram, discuss how our field analysis influenced our design decisions, and present a qualitative, long-term, in-depth evaluation. Results of this study showed that our participants successfully used Cardiogram to increase the amount of analyzable information, to externalize domain knowledge, and to provide new insights into trace data. Our design approach finally led to the adoption of Cardiogram into engineers' daily practices.