The implementation of social policies is often hampered by the fact that local decision-makers may be unwilling or unable to implement the policy as intended by the central government. In contrast to research that focuses on incentivizing and holding local decision makers accountable, we examine capacity constraints in the context of beneficiary selection. Using a large-scale randomized trial in Bangladesh, we find that training and data provision improved knowledge of selection criteria. However, evidence of better targeting was limited, except for easily observable indicators of vulnerability. Improvements in targeting were more pronounced in committees led by highly educated chairpersons.