The impact of a blended ICT adoption model in 15 fourth grade classes from 11 vulnerable Chilean schools is analyzed. In this model, twice a week students attend a computer lab where the lab teacher and the class teacher select and assign on line math exercises. The platform contains approximately 2,000 exercises, but it is constantly growing with exercises introduced by teachers. During sessions the system continuously tracks students and detects students that are falling behind, allowing teachers to provide real time support and drive the progression of the entire class as a whole. Students that finish early and with good performances are assigned as part of the support team for the rest of the session. After a year of implementation, the national assessment test SIMCE math on these classes raised 0.38 standard deviations. This is more than three times the historic national improvement in 2011. The statistical analysis shows that the improvement was independent of teacher effects, and correlates with the average number of on line exercises done per student in that year.