In this article we test whether metrics of online attention describing research can provide information on research quality and societal impact that is not found in citation data alone. Our approach is to set up a traditional model in which the true quality or impact of a university department is determined by a panel of experts, but a citation metric is regarded as a reasonable proxy. However, the model assumes that the information contained in the scores provided by an expert panel exceeds that contained in a citation metric (HEFCE, 2015). Finally, we extend this model by including altmetric data to see if it adds information about a department's performance that cannot be gleaned from citations alone. We find the presence of altmetric data for the cited underpinning research to be highly correlated with peer review scores for societal impact. Conversely, no such connection was seen with the assessment of research quality. Our findings therefore suggest altmetric data could be useful as an aid to assessing impact. Data SourcesThe analysis was restricted to main panel B of the submissions, which relates to mathematics, engineering, and the physical sciences (see Table 1). Within this subject grouping, 48,815 of the submitted 49,317 research outputs were journal articles, allowing a near complete mapping of bibliometric indicators to REF scores. The research activity in this panel is subdivided into nine separate topics; however, for the purpose of this work, three of the engineering units of assessments (UOAs) were combined, as they could all be labeled as "applied engineering." The creation of this agglomerated group meant that the number of submitted outputs across the subject areas were somewhat equalized