Requirements for outcomes–based performance management are increasing performance–evaluation activities at all government levels. Research on public–sector performance management, however, points to problems in the design and management of these systems and questions their effectiveness as policy tools for increasing governmental accountability. In this article, I analyze experimental data and the performance–management experiences of federal job–training programs to estimate the influence of public management and system–design factors on program outcomes and impacts. I assess whether relying on administrative data to measure program (rather than impacts) produces information that might misdirect program managers in their performance–management activities. While the results of empirical analyses confirm that the use of administrative data in performance management is unlikely to produce accurate estimates of true program impacts, they also suggest these data can still generate useful information for public managers about policy levers that can be manipulated to improve organizational performance.