Social robots are designed to interact with humans. That is why they need interaction models that take social behaviors into account. These usually influence many of a robot’s abilities simultaneously. Hence, when designing robots that users will want to interact with, all components need to be tested in the system context, with real users and real tasks in real interactions. This requires methods that link the analysis of the robot’s internal computations within and between components (system level) with the interplay between robot and user (interaction level). This article presents Systemic Interaction Analysis (SInA) as an integrated method to (a) derive prototypical courses of interaction based on system and interaction level, (b) identify deviations from these, (c) infer the causes of deviations by analyzing the system’s operational sequences, and (d) improve the robot iteratively by adjusting models and implementations.