Abstract. Models defined using the Unified Modeling Language (UML) are nowadays common parts of software documentations, specifications and sometimes even implementations. However, there is a broad variety of how UML is used. Reasons can be found, for example, in the lack of generally accepted modeling norms and guidelines, the semi-formal semantics of UML, or the complexity of the language. In practice, these factors inevitably lead to quality problems in UML models that need to be addressed. We investigate and discuss existing work in the field of quality assessment and improvement of UML models and present how we envision an integrated approach to quality assessment and improvement of UML models. We assess a model with a Factor-Criteria-Metrics (FCM) based quality model, detect issues by finding smells and violated metric thresholds in UML models, and improve UML models by applying refactorings using model-to-model transformations.