Quality certification is a common tool to reduce asymmetric information and enhance trust in marketplaces. Should the certificate focus on seller inputs such as fast shipping, or include output measures such as consumer ratings? In theory, incorporating output measures makes the certificate more relevant for consumer experience, but doing so may discourage seller effort because outputs can be driven by random factors out of seller control. To understand this tradeoff, we study a major redesign of eBay's Top Rated Seller (eTRS) program in 2016, which removed most consumer reports from the eTRS criteria and added direct measures of seller inputs. This change generates immediate selection on certified sellers, and homogenizes the share of certified sellers across product categories of different criticalness in consumer ratings. Post the regime change, sellers improve in the input measures highlighted in the new certificate. These effects are more conspicuous in categories that had less critical consumer ratings, in part because the new algorithm automatically removes the potential negative bias for sellers in critical markets and a clearer threshold motivates sellers to just reach the threshold. The new regime also makes sales more concentrated towards large sellers, especially in the categories that face more critical consumers.