Providing financial incentives to encourage behavioral change is increasingly common: policymakers pay people to exercise, to study more, and maintain tree cover. In designing incentives for behavior change, there is frequently a tradeoff between motivating those with high and low costs of engaging in the behavior: whereas the optimal contract for high-cost individuals will have a large incentive or a low behavior target, the optimal contract for low-cost individuals will have a small incentive or a large behavior target. This project will investigate methods for targeting incentive contracts for behavioral change to individuals with heterogeneous behavior cost. In a randomized controlled trial among individuals with diabetes and prediabetes, we will experimentally evaluate two methods for targeting incentive contracts for walking: individual contract choice, and targeting on observable characteristics. We will first show that in our setting, contracts with a higher step target are more effective for individuals with low walking costs at baseline. Second, we will assess the relative performance of our two targeting methods in terms of targeting precision and walking encouragement. Finally, we will explore the role of key challenges to implementing each method, including the endogeneity and limited information content of observable characteristics, and the principal and agent’s imperfect knowledge of preferences.