Using a student level randomization, we compare three education-based conditional cash transfers designs: a standard design, a design where part of the monthly transfers are postponed until children have to re-enroll in school, and a design that lowers the reward for attendance but incentivizes graduation and tertiary enrollment. The two nonstandard designs significantly increase enrollment rates at both the secondary and tertiary levels while delivering the same attendance gains as the standard design. Postponing some of the attendance transfers to the time of re-enrollment appears particularly effective for the most at-risk children. (JEL H23, I21, I22, J13, O15)