Purpose: The purpose of this study was to design and iteratively improve the quality of survey-based measures of three non-cognitive constructs for Grade 5-6 students, keeping in mind information needs of users in education reform contexts. The constructs are: Mathematics-related Self-Efficacy, Self-Concept, and Anxiety (M-SE, M-SC, and M-ANX). Design/methodology/approach: The authors applied a multi-stage, iterative and user-centered approach to design and validate the measures, using several psychometric techniques and three data samples. They evaluated the utility of student-level scores and aggregated, classroom-level means. Findings: At both student and classroom levels, replicated evidence supported theoretically-grounded validity arguments on information produced by four of five scales tapping M-SC, M-ANX and M-SE. The evidence confirmed a second order, two-factor structure for M-SC, representing positive math affect and perceived competence, and a one factor structure for M-ANX representing negative math affect. Consistent with the literature, these served as precursors to a perceived confidence factor of M-SE which, in turn, positively influenced mathematics achievement scores, off-setting negative effects of M-ANX. Research is continuing on a self-regulatory efficacy factor of M-SE, which yielded mixed results. Practical implications: The survey scales are in line with current reform policies in the United States calling for schools to monitor changes in cognitive and non-cognitive domains of student development. Validated scales could be useful in serving information needs of teachers, decision-makers and researchers in similar school-based contexts. Originality/value: This study demonstrates a comprehensive, user-centered methodology for designing and validating construct measures, departing from purely psychometric traditions of scale development.