Realist review has increased in popularity as a methodology for complex intervention assessment. Our experience suggests that the process of designing a realist review requires its customization to areas under investigation. To elaborate on this idea, we first describe the logic underpinning realist review and then present critical reflections on our application experience, organized in seven areas. These are the following: (1) the challenge of identifying middle range theory; (2) addressing heterogeneity and lack of conceptual clarity; (3) the challenge of appraising the quality of complex evidence; (4) the relevance of capturing unintended outcomes; (5) understanding the process of context, mechanism, and outcome (CMO) configuring; (6) incorporating middle-range theory in the CMO configuration process; and (7) using middle range theory to advance the conceptualization of outcomes – both visible and seemingly ‘hidden’. One conclusion from our experience is that the degree of heterogeneity of the evidence base will determine whether theory can drive the development of review protocols from the outset, or will follow only after an intense period of data immersion. We hope that presenting a critical reflection on customizing realist review will convey how the methodology can be tailored to the often complex and idiosyncratic features of health research, leading to innovative evidence syntheses.