Evaluations that use imperfect information run into both analytical and political problems. Educational accountability systems based on achievement scores are an instance. Such systems frequently turn out to be irrelevant to policy decisions, resisted by educational interest groups that fear unflattering comparisons and the misuse of results, and infeasible. This paper suggests several ways to analyze limited information that both improve the decisionmaking relevance of evaluation and decrease the adverse political consequences. The context for discussing these issues is a case study of an early stage in the design of a State accountability system for public schools. The paper discusses a series of theoretical and practical problems in assessing the effects of school policies with imperfect achievement data and limited controls for nonschool factors that influence scores. Problems of the collection, analysis, and presentation of data are explored in the case study. Hypothetical regression equations and tables of policy-relevant results are provided as illustrations. (Author/WM)