This paper focuses on how educational statistics might better serve the quest for educational improvement in elementary and secondary schools. A model for conceptualizing the sources and processes of school productivity is presented. The Learning Productivity Model suggests that school outcomes are the consequence of the interaction of five factors: (1) clarity of educational objectives; (2) student characteristics; (3) curriculum technology and instructional strategies; (4) teacher quality; and (5) school environment--opportunity and resources. The importance of educational research is stressed for its ability to identify what types of data are relevant in studying effective schools. The model considers a variety of educational outcomes, both academic achievement and lifetime outcomes or post-school experiences such as occupational success. Two suggestions for data collection are discussed. New data should be used to relate educational processes to educational outcomes, as in the High School and Beyond study. Better coordination of data collection activities is recommended. (GDC)