What goes on inside the classroom is central to student learning. Despite its importance, low- and middle-income countries rarely measure teaching practices, in part due to a lack of access to adequate classroom observation tools and the high transaction costs associated with administering them. Teach, a new, open-source classroom observation tool for primary classrooms, was developed to capture the quantity and quality of teaching practices in these settings with a simple, easy-to-administer tool. This paper validates the use of Teach scores for system diagnostics by providing four types of evidence. First, it provides evidence that the practices included in the tool have a clear conceptual underpinning. Second, almost 90 percent of local observers in Mozambique, Pakistan, the Philippines, and Uruguay were highly accurate using Teach after a four-day training. Third, using data from 845 classrooms in Pakistan, the paper shows that Teach scores are internally consistent, present moderate to high inter-rater reliability in the field (.75 intraclass correlation coefficient), and provide substantial information that allows to differentiate teachers, even those with similar but not equal scores. Finally, teachers who display effective practices, as measured by Teach, are associated with students who achieve higher learning outcomes.