Conditional Cash Transfer (CCT) programs have become extensively used to induce poor parents to increase their investments in the human capital of their children. The condition on school attendance and use of health facilities transforms the transfer into a price effect on the condition. Justification for the condition is to reduce market failures due to positive externalities from investments in human capital, while transferring money to the poor. To be efficient, CCT programs thus need to successfully implement three rules. The first is a rule to select the poor. The other two are rules of eligibility among the poor and of calibration of transfers, particularly if budgets are insufficient to offer large universal transfers to all the poor. Using the case of Progresa in Mexico, we show that efficiency gains can be achieved by taking into account the probability of enrollment of a child, and how it is expected to respond to a cash transfer. Calibration relies on heterogeneity in responses due to child, household, and community characteristics. Rules can be made easily implementable by selecting indicators that are simple, easily observable and verifiable, and that cannot be manipulated by beneficiaries. We show that, when programs operate under strong budgets constraints, major efficiency gains can indeed be achieved by careful design of eligibility and transfer rules. In the case under study, these efficiency gains can be achieved without equity costs on the poor.