Purpose: Though test-based accountability policies seek to redress educational inequities, their underlying theories of action treat inequality as a technical problem rather than a political one: data point educators toward ameliorative actions without forcing them to confront systemic inequities that contribute to achievement disparities. To highlight the problematic nature of this tension, the purpose of this paper is to identify key problems with the techno-rational logic of accountability policies and reflect on the ways in which they influence teachers' data-use practices. Design/methodology/approach: This paper illustrates the data use practices of a workgroup of sixth-grade math educators. Their meeting represents a "best case" of commonplace practice: during a full-day professional development session, they used data from a standardized district benchmark assessment with support from an expert instructional leader. This sociolinguistic analysis examines episodes of data reasoning to understand the links between the educators' interpretations and instructional decisions. Findings: This paper identifies three primary issues with test-based accountability policies: reducing complex constructs to quantitative variables, valuing remediation over instructional improvement, and enacting faith in instrument validity. At the same time, possibilities for equitable instruction were foreclosed, as teachers analyzed data in ways that gave little consideration of students' cultural identities or funds of knowledge. Social implications: Test-based accountability policies do not compel educators to use data to address the deeper issues of equity, thereby inadvertently reinforcing biased systems and positioning students from marginalized backgrounds at an educational disadvantage. Originality/value: This paper fulfills a need to critically examine the ways in which test-based accountability policies influence educators' data-use practices.