We examine optimal incentives for charitable giving with a large-scale field experiment involving 26 charities and over 112,000 unique individuals. The price of giving is varied by offering a fixed match if the donation meets a threshold amount (e.g. "give at least $25 and the charity receives a $25 match"). Responses are used to structurally estimate a model of charitable giving. The model estimates are employed to evaluate the effectiveness of various counterfactual match incentive schemes, taking into account the goals of the charity and donor preferences. Two of these optimal incentives were subsequently implemented in a follow-up field study. They were found to be effective at implementing the desired goals, as predicted by theory and our simulations. Our findings highlight the pitfalls of relying on a particular parameterization of a policy to evaluate effectiveness. The best-guess incentives in our initial field experiment turned out to be ineffective at increasing donations because optimal incentives should have been set higher.