Teacher coaching is a powerful form of professional learning that improves teaching practices and student achievement, yet little is known about the specific aspects of coaching programs that are more effective. Researchers used a blocked randomized experiment to study the effects of one-to-one coaching on teacher practice. When pooled across all teachers in both cohorts, there is no effect of coaching on teacher practice, yet considerable variability exists between the cohorts. Changes in program design that occurred between the two cohorts provided researchers an opportunity to study how differences in program features influence positive effects in the first cohort on teacher practice and the absence of effects in the second cohort.