An increasing number of metrics are used to measure the impact of research papers. Despite being the most commonly used, the 2-year impact factor is limited by a lack of generalisability and comparability, in part due to substantial variation within and between fields. Similar limitations apply to metrics such as citations per paper. New approaches compare a paper’s citation count to others in the research area, while others measure social and traditional media impact. However, none of these measures take into account an individual author’s contribution to the paper or the number of authors, which we argue are key limitations. The UK’s 2014 Research Exercise Framework included a detailed bibliometric analysis comparing 15 selected metrics to a ‘gold standard’ evaluation of almost 150 000 papers by expert panels. We outline the main correlations between the most highly regarded papers by the expert panel in the Psychiatry, Clinical Psychology and Neurology unit and these metrics, most of which were weak to moderate. The strongest correlation was with the SCImago Journal Rank, a variant of the journal impact factor, while the amount of Twitter activity showed no correlation. We suggest that an aggregate measure combining journal metrics, field-standardised citation data and alternative metrics, including weighting or colour-coding of individual papers to account for author contribution, could provide more clarity.