In a common approach for scaling up effective educational practice, schools adopt evidence-based programs to be implemented with fidelity. An alternative approach assumes that programs should be adapted to local contexts. In this randomized trial of a reading intervention, we study a scaffolded sequence of implementation in which schools first develop proficiency by implementing the program with fidelity before implementing structured adaptations. We find evidence supporting the scaffolded sequence: A fidelity-focused approach promoted learning and instructional change more so for teachers inexperienced with the intervention, while a structured adaptive approach was more effective for teachers experienced with the intervention. Students benefited more from the structured adaptive approach but only when their teacher had prior experience with the fidelity-focused version.