This qualitative, empirical study explores and describes the variation in how evidence-based practice (EBP) is understood in social work. A phenomenographic approach to design and analysis was applied. Fourteen semi-structured interviews were conducted with politicians, managers and executive staff in three social welfare offices in Sweden. The main findings suggest that there are qualitatively different ways in which EBP is understood, described in five categories: (i) fragmented; (ii) discursive; (iii) instrumental; (iv) multifaceted; and (v) critical. The outcome space is hierarchically structured with a logical relationship between the categories. However, the informants found it difficult to account for EBP, depending on what was expressed as deficient knowledge of EBP in the organisation, as well as ability to provide a seemly context for EBP. The results highlight the importance of acknowledging these differences in the organisation to compose a supportive atmosphere for EBP to thrive rather than merely assume the case of evidence-based social work. The categories can be utilised as stimuli for reflection in social work practice, and thereby provide the possibility to promote knowledge use and learning in the evolving evidence-based social work.