The No Child Left Behind (NCLB) Act required states to adopt accountability systems measuring student proficiency on state-administered exams. The federal legislation contained several strict requirements for NCLB implementation, such as escalating student proficiency targets that reach 100% proficiency by 2014. But it also gave states considerable flexibility to interpret and implement components of NCLB. Using a data set we constructed, this paper is the first national study examining which schools failed during the early years of NCLB and which performance targets they failed to meet. We explore how states' NCLB implementation decisions were related to their schools' failure rates, which ranged from less than 1% to more than 80% across states. Wide cross-state variation in failure rates resulted from how states' decisions interacted with each other and with school characteristics, like enrollment size, grade span, and ethnic diversity. Subtle differences in policy implementation may cause dramatic differences in measured outcomes.