Despite recent shifts in research emphasizing the value of carefully designed experiments, the number of studies of teacher professional development with rigorous designs has lagged behind its student outcome counterparts. We outline a framework for the design of group randomized trials (GRTs) with teachers' knowledge as the outcome and consider mathematics and reading knowledge outcomes designed to assess the types of content problems that teachers encounter in practice. To estimate design parameters, we draw on a national sample of teachers for mathematics and a state Reading First sample to estimate for reading. Our results suggest that there is substantial clustering of teachers' knowledge within schools and professional development GRTs will likely need increased sample sizes to account for this clustering. (Contains 1 note, 9 tables, and 2 figures.)