Teacher observations have become a national education phenomenon, mandated by federal policies and promoted by philanthropists. They are crucial components of teacher evaluation systems that often have high stakes for teachers and school systems, but have sparked little innovation. Recent calls to make teacher evaluations better, faster, and cheaper have been challenged as sitting outside the research evidence. In response, we present new evidence from experimental studies of the Rapid Assessment of Teacher Effectiveness (RATE) that suggests that innovation of that kind is possible. Using a six-item rubric after 4 hr of training, raters were able to identify effective teachers from just 20 min of one lesson as well or better than raters using popular evidence-based instruments with 10 times the items. We discuss our approach, the potential time and cost savings it may provide, and how better, faster, cheaper observation instruments may change how school districts choose to implement state and federal policies.