This article highlights the development process and data-driven decisions that characterized a 5-year iterative process of development and evaluation of a Tier 2 early language curriculum, Story Friends. This curriculum was specifically designed to be effective and feasible for high fidelity implementation in preschool settings. Vocabulary and comprehension lessons were embedded within storybooks. The stories and lessons were prerecorded so adults could easily administer the intervention to groups of three or four children. The development and selection of stories, intervention targets, instructional components, measurement schemes, and training materials were improved in response to data and feedback from an array of implementers. A series of research designs were used as the intervention was scaled up as warranted by efficacy results. The process of iterative development and refinement described in this article is one that is needed to avoid premature and expensive evaluations of underdeveloped treatments that too often produce weak intervention effects. (author abstract)