Teachers are now experiencing the age of quantitative test-driven assessment, in which there is little weight accorded to teacher-based judgement about student progress. In the Australian context, the NAPLaN test has become a driving force in school and teacher accountability. The language of NAPLaN is one of bands and numerical scores and comparative performance, measuring schools against other "similar" schools and against a national average. The consequences of this are troubling for all teachers. For EAL/D teachers whose specialised professional knowledge relates to building the academic English language of EAL/D learners, NAPLaN is highly problematic because it takes no account of second language factors which might impact on test performance. Yet, NAPLaN data do offer rich yet largely unexploited opportunity to highlight the validity of teacher judgement in the classroom. This paper will use ESL Bandscale data in an analysis of the NAPLaN performance of EAL/D students to show how teacher judgement (captured by the ESL Bandscales) is valid and aligned with NAPLaN performance. I will demystify the power and the fallibility of large-scale assessment like NAPLaN: to identify in which contexts large-scale data analysis is useful, and its limitations in micro-settings which include the classroom. The second goal of this paper is to stress the utility of teacher data, measured quantitatively, but based on qualitative observation, when it is grounded in teacher professional knowledge. Ultimately, such arguments serve to highlight the importance of remaining grounded in strong TESOL pedagogy, despite the intense pressure to follow mainstream English (as first language) literacy responses to NAPLaN testing.