Educational interventions typically are complex combinations of human actions, organizational supports, and instructional resources that play out differently in different contexts and with different kinds of students. The complexity and variability of outcomes undermines the notion that interventions either "work" or "don't work." Under the design-based implementation research (DBIR) model, the implementation of an intervention in particular settings is itself an object of research and a critical part of understanding how to scale an intervention without diluting its effectiveness. In this chapter, we compare the approach to evidence implicit in the defining features of DBIR to the prevailing evidence standards for educational research promoted by national policy. Our aim is to provide a frame for knowledge building within DBIR that draws from the strengths of both design-based research methods and research designs that permit causal inference about program impacts. Moreover, we endeavor to show how DBIR challenges current thinking about what counts as credible research. We conclude by considering the ways in which DBIR is a departure from much educational research in terms of what it means to conduct research that is useful and usable in education settings.