Input subsidy programs carry support as instruments to increase agricultural productivity, provided they are market-smart. This requires especially proper targeting to contain the fiscal pressure, with decentralized targeting of input vouchers currently the instrument of choice. Nonetheless, despite clear advantages in administrative costs, the fear of elite capture persists. These fears are borne out in the experience from the 2008 input voucher pilot program in Kilimanjaro, Tanzania, examined here. Elected village officials received about 60 percent of the distributed vouchers, a factor that significantly reduced the targeting performance of the program, especially in more unequal and remote communities. When targeting the poor, greater coverage and a focus on high trust settings helped mitigate these concerns. The findings highlight the continuing need for scrutiny when relying on decentralized targeting. A clearer sense of purpose (increasing productivity among poorer farmers versus increasing aggregate output) could also enhance the targeting performance.