Progress in understanding the underlying neurobiology of obsessive-compulsive disorder (OCD) has stalled in part because of the considerable problem of heterogeneity within this diagnostic category, and homogeneity across other putatively discrete, diagnostic categories. As psychiatry begins to recognize the shortcomings of a purely symptom-based psychiatric nosology, new data-driven approaches have begun to be utilized with the goal of solving these problems: specifically, identifying trans-diagnostic aspects of clinical phenomenology based on their association with neurobiological processes. In this review, we describe key methodological approaches to understanding OCD from this perspective and highlight the candidate traits that have already been identified as a result of these early endeavours. We discuss how important inferences can be made from pre-existing case-control studies as well as showcasing newer methods that rely on large general population datasets to refine and validate psychiatric phenotypes. As exemplars, we take ‘compulsivity’ and ‘anxiety’, putatively trans-diagnostic symptom dimensions that are linked to well-defined neurobiological mechanisms, goal-directed learning and error-related negativity, respectively. We argue that the identification of biologically valid, more homogeneous, dimensions such as these provides renewed optimism for identifying reliable genetic contributions to OCD and other disorders, improving animal models and critically, provides a path towards a future of more targeted psychiatric treatments.