Excellence policies in science are becoming increasingly widespread for research quality assessment in a context of growing public accountability of broader social, cultural, environmental or economic returns to society of publicly funded research. Existing approaches to research quality assessment, however, insufficiently acknowledge the heterogeneity of individual scientists’ profiles which make scientific research conducive to major contributions for society through many different pathways, involving a wide diversity of actors, outputs, and long-term impacts. In this paper we first critically discuss the concept of excellence in research assessment drawing on the experience of the Research Excellence Framework (REF) in the UK. Rather than a clear focus on denoting what is excellent in scientific research, we find that the REF enables the identification of diverse contributions to top quality research and impacts by UK departments and units. Then we propose a model for better understanding the production of research quality. Like the UK REF, our model features scientific outputs and social impacts as the two major dimensions of research quality for individual research assessment. We draw on the individual scientist as a point of reference to better understand the individual and contextual pre-conditions that generate pathways to social impact through productive interactions. The ultimate objective is to produce insights into the diversity of pathways and the complementarities between producing scientific and social impacts.