To identify genetic risk loci for major depressive disorder (MDD), two broad study design approaches have been applied: (1) to maximize sample size by combining data from different phenotype assessment modalities (e.g. clinical interview, self-report questionnaires) and (2) to reduce phenotypic heterogeneity through selecting more homogenous MDD subtypes. The value of these strategies has been debated. In this review, we summarize the most recent findings of large genomic studies that applied these approaches, and we highlight the merits and pitfalls of both approaches with particular attention to methodological and psychometric issues. We also discuss the results of analyses that investigated the heterogeneity of MDD. We conclude that both study designs are essential for further research. So far, increasing sample size has led to the identification of a relatively high number of genomic loci linked to depression. However, part of the identified variants may be related to a phenotype common to internalizing disorders and related traits. As such, samples containing detailed clinical information are needed to dissect depression heterogeneity and enable the potential identification of variants specific to a more restricted MDD phenotype. A balanced portfolio reconciling both study design approaches is the optimal approach to progress further in unraveling the genetic architecture of depression.