This paper studies how design features influence the success of Housing Mobility Programs (HMPs) in reducing racial segregation. Targeting neighborhoods based on previous residents' outcomes does not allow for targeting race-specific outcomes, generates uncertainty when targeting income-specific outcomes, and generates bias in ranking neighborhoods' effects. Moreover, targeting opportunity bargains based on previous residents' outcomes selects tracts with large disagreements in current and previous residents' outcomes, with such disagreements predicted by sorting since 1990. HMP success is aided by the ability to port vouchers across jurisdictions, access to cars, and relaxing supply constraints, perhaps by targeting lower-ranked neighborhoods.