To identify and correct faults in a technical system, people have to understand the underlying cause-effect relations and apply this knowledge to derive appropriate actions. Causal diagrams have previously been found to enhance people’s understanding of direct and indirect causal influences. However, these diagrams are typically very small and not representative of causal relations in complex systems. How do people use large causal diagrams to identify fault causes and to draw more abstract inferences from the provided relations? What do they learn about the represented contents? To answer these questions, we conducted an experiment in which participants used a complex causal diagram (88 variables) representing cause-effect relations during chocolate production. They were asked to identify potential fault causes, and to reason about more abstract principles underlying different faults. To assess learning, participants were also asked to identify potential problems without the causal diagram. Using participants’ written records and think-aloud protocols, we analyzed their strategies and challenges. Participants identified most causal relations correctly, but contrary to previous research, they often failed to identify long causal chains and indirect paths. About one third of participants were not able to derive process-related abstract principles underlying the causal relations. Finally, the results indicate that participants have learned the cause-effect relations on very low and on very high levels of functional abstraction, which are often unsuitable to derive appropriate actions. We discuss potential adaptations to address these shortcomings, and how benefits and limitations of large causal diagrams could be examined in future studies.