Many educational interventions seek to change teachers' instructional practice. Standards-based observation systems are a common and useful tool to understand these interventions' impact, but the process of measuring instructional change with observation systems is highly complex. This paper introduces a framework for examining and understanding potential instrumentation biases that arise when evaluations use observation systems to understand instructional change. The framework systematizes two processes that all studies must undertake: (1) the process of operationalizing the construct of teaching quality, and (2) the process of data collection. A study that engages in these processes generates observation scores that capture their own raters' perspectives on specific segments of instruction. These scores must be generalized to draw conclusions about the intended constructs and settings. Systematizing these two processes highlights the necessary steps of a validity argument supporting evaluation conclusions and the instrumentation biases that threaten such conclusions. The framework is illustrated with an example from our recent work, which sought to understand instructional change since the adoption of the Common Core State Standards (CCSS).