This paper offers a systematic exploration of online feedback mechanism design issues in trading environments with opportunistic sellers, imperfect monitoring of a seller's effort level, and two possible transaction outcomes (corresponding to "high" and "low" quality respectively), one of which has no value to buyers. The objective of feedback mechanisms in such settings is to induce sellers to exert high effort and, therefore, to maximize the probability of high quality outcomes. I study a practically significant family of mechanisms that resembles aspects of the one used by online auction house eBay. These feedback mechanisms solicit "binary" ratings of transaction outcomes as either positive or negative and publish the sums of ratings posted by buyers on a seller during the N most recent periods. My analysis finds that such "binary" feedback mechanisms can induce high average levels of cooperation that remain stable over time. Surprisingly, their efficiency cannot be improved by summarizing larger numbers of ratings or by publishing a seller's detailed feedback history. I further examine the robustness of these mechanisms to incorrect or incomplete feedback as well as to strategic changes of online identities. The theoretical outcomes predicted by this paper are consistent with empirical observations and offer theory-backed explanations to hitherto poorly understood phenomena such as the remarkably low fraction of negative feedback on eBa