The paper evaluates math performance at four high-need middle schools during a four-year intervention, which was designed to help math teachers diagnose students' areas of need and to design lesson plans responsive to those needs. Before the intervention began, the researchers pre-selected four comparison schools by matching based on achievement and also on demographics. A difference-in-difference analysis finds a significant increase of about 0.11 standard deviation in test scores per year for students in the program schools. Supplementary event study and synthetic control analyses to detect year-by-year effects lack precision but are weakly suggestive of a smaller impact in year 1 than later years. A cost analysis considers the affordability of extending similar programs.