Commitment devices offer an opportunity to restrict future choices. However, if severe restrictions deter participation, weaker restrictions may be a more effective means of changing behavior. We test this using a school-based commitment savings device for educational expenses in Uganda. We compare an account fully-committed to educational expenses to an account in which savings are available for cash withdrawal but intended for educational expenses. The weaker commitment generates increased savings in the program accounts and when combined with a parent outreach program, higher expenditures on educational supplies. It also increases scores on an exam covering language and math skills by 0.14 standard deviations. We find no effect for the fully-committed account, and we find no effect for either account on attendance, enrollment, or non-cognitive skills.