Intervention models based on data use can be effective in raising student achievement. This article presents 3 studies of one such model which had reported improved reading comprehension levels in 7 poor urban multicultural schools serving indigenous and ethnic minority communities. The intervention (the Learning Schools Model) used a process comprising critical discussions of achievement and teacher observation data to develop specific and contextualized content for fine-tune instruction. The reliability and generality of the effects of the model were tested in a cluster of "like" schools and a cluster of "unlike" schools. The growth models showed similar effects to the original schools, with gains of between 3 to 4 months additional progress per year over 3 years. The replications show that models that use data to design local program content can be reliably and generally effective, but also there is a need to examine differential effects. (Contains 8 figures, 6 tables, and 1 note.)