Accelerators are an increasingly prevalent and important part of the entrepreneurial landscape. They provide intensive mentoring and education over a short, fixed-length period with the goal of aiding early stage ventures’ development. Yet, while recent research indicates that some accelerators have a causal treatment effect, questions remain about why some accelerators seem to have a greater effect than others and some have no effect or even a negative effect. We address this gap using an inductive, theory-building approach that mobilizes rich fieldwork within the cohorts of eight early US accelerator programs. Our data suggest three practices that differentiate accelerator effectiveness. Collectively these practices enable some accelerators to effectively mitigate problems arising from the bounded rationality of entrepreneurs. We discuss implications for theory and practice.