This paper investigates the short term effects of a large scale intervention, funded by the European Social Fund, that provides additional instruction time to students in low achieving lower secondary schools of Southern Italy. We control for sorting across classes using the fact that freshman are divided into groups distinguished by letters, they remain in the same group across grades and the composition of teachers in the school assigned to each group is substantially stable over time. We implement a difference-in-differences strategy, and compare contiguous cohorts of freshman enrolled in the same group. We contrast groups with and without additional instruction time in participating schools, to groups in non-participating schools that we select to be similar with respect to a long list of pre-programme indicators. We find that the programme raised test scores in mathematics in schools characterised by students from less advantaged backgrounds. We also find that targeting the best students with extra activities in Italian language comes at the cost of lowering their performance in mathematics. We go beyond average effects, finding that the positive effect documented for mathematics is driven by larger effects for the best students in the group.