THE PRIMARY PURPOSE OF EVALUATION IS UTILITARIAN. IT CAN SERVE OTHER FUNCTIONS AS WELL, SUCH AS TESTING THEORY, BUT ITS BASIC FUNCTION IS TO PROVIDE EVIDENCE ABOUT PROGRAM EFFECTIVENESS. PROBLEMS IN TRYING TO COMBINE POLICY AND ADMINISTRATIVE EVALUATION RELATE TO RESEARCH VALIDITY, THE NEED TO EXPLICITLY DEFINE AND FIRMLY CONTROL A PROGRAM, AND THE SCOPE AND COST OF EVALUATION. IDEAL REQUIREMENTS FOR EVALUATION IN THE POLICY CONTEXT ARE AN EXPERIMENTAL DESIGN, AN EXPLICIT AND STABLE PROGRAM, AND SUFFICIENT SITES FOR GENERALIZABILITY. SOCIAL EXPERIMENTS, HOWEVER, ARE COSTLY BECAUSE A PROTOTYPE PROGRAM MUST BE ADMINISTERED AS COMPLETELY AS A SOPHISTICATED RESEARCH STUDY. SOCIAL EXPERIMENTS LOOK TOWARD FUTURE PLANS, BUT POLICY DECISIONS MUST BE MADE ABOUT EXISTING PROGRAMS. FOR ADMINISTRATIVE PURPOSES, TRADITIONAL EVALUATION METHODS HAVE CONSIDERABLE POTENTIAL IF THEY ARE WELL EXECUTED. INFORMATION SYSTEMS ARE USEFUL IN ADMINISTRATION AND CAN BE EXPANDED TO SERVE ACCOUNTABILITY FUNCTIONS. ONCE RELEVANT AUDIENCES ARE IDENTIFIED AND THEIR CRITERIA OF PROGRAM EFFECTIVENESS ARE UNDERSTOOD, EVALUATORS CAN DEVELOP MEASURES OF THESE CRITERIA AND PUT THEM INTO INFORMATION SYSTEMS. WITH OUTCOME DATA DEFINED BY RELEVANT USER GROUPS, THE ACCOUNTABILITY FUNCTION ALLOWS PERIODIC REPORTING TO EACH GROUP. GOALS IN EVALUATION DESIGN TO MEET THE NEEDS OF DECISIONMAKERS ARE ESSENTIAL. (DEP)