We study how attendance rates of primary school children respond to cost neutral changes in the design of India's school meal program. Municipal schools in the capital region of Delhi switched from packaged food to on-site cooked meals in 2003, with insignificant changes in the budget available per meal. We use the staggered implementation of this transition and child-level panel data to find that cooked meals resulted in a 3-4 percentage point rise in average monthly attendance with the largest effects observed for early grades. The impact on girls was especially large, but since they attend morning shift schools while boys attend the afternoon shifts, these gender effects may simply reflect benefits from better timed meals. We also find attendance gains concentrated in schools that served diverse menus. Our results suggest that even within constrained budgets, better designed school transfer programs can improve student level outcomes.