A common theme running through recent research on Situational Judgment Tests (SJTs) and this special issue is the aim to improve the measurement of constructs via SJTs. Construct-driven SJTs differ from traditional SJTs in that they present a trait activating situation to test-takers and a more unidimensional set of response options that depict different trait levels. In this commentary, I frame the different papers of this special issue into a research agenda related to construct-driven SJTs. Specifically, I posit that future research should examine whether construct-driven SJTs lead to more unidimensionality at the item level, cleaner measurement of the constructs, and more equivalence of SJT scores across settings. In addition, the effects of using a construct-driven approach on the criterion-related validity of SJT scores and their convergence with actual behavior need to be scrutinized. Finally, I suggest investigating the susceptibility of construct-driven SJTs to faking, retest, item exposure, and coaching effects.