The last two decades have witnessed a vast expansion in the use of education data to improve classroom instruction and raise student achievement. Schools and districts face important challenges in implementing increased data use for instructional improvement. One key challenge is the need for teachers and administrators to have "data literacy"--the skills to analyze data, and to use a variety of data sources to refine and improve instruction. Data systems and data initiatives have grown at a much faster pace than educator training around data use. This reality justifies the evaluation of a program such as TERC's Using Data, which aims to provide teachers with the needed training. A table and figure are appended. Using school-level random assignment, this study seeks to estimate the causal impact of TERC's Using Data (UD) program on elementary teachers' understanding and use of data to inform mathematics instruction, and on the mathematics achievement of their fourth and fifth grade students. The main questions were: (1) At the end of the first year of the intervention, do fourth and fifth grade students in schools randomly assigned to the UD treatment program show greater levels of math achievement as compared to their counterparts in control schools? (2) After the first year of the intervention, do UD-trained teachers in treatment schools report greater data-use knowledge and skills and more-positive attitudes and beliefs about the value of data to inform instruction as compared to their same-grade counterparts in control schools?; and (3) After the first year of the intervention, do UD-trained teachers in treatment schools reportedly make more use of data or work more frequently with data in a collaborative setting, as compared to their same-grade counterparts in control schools? in comparison to control teachers? A large, urban district in the southeastern U. S. joined the study in March 2011, agreeing to implement the Using Data program. Sixty schools were randomized into the treatment group (30 schools) or the control group (30 schools), with an average of four teacher participants and one data coach from each school taking on the same treatment condition as their school. Our sample includes over 11,000 students and 800 math teachers in the 60 schools. The design was a block-randomized experiment with two levels of treatment--Using Data versus practice as usual. Preliminary analysis of pre/post survey data from teachers suggests that collaborative data use increased in the treatment schools relative to the control schools at statistically significant levels. Tables and one figure are appended.