'METAEVALUATION' REFERS TO VARIOUS ATTEMPTS TO JUDGE THE MERIT OF EVALUATION WORK AND ITS IMPACT ON MANAGERIAL DECISIONS. DECISION IMPACT IS THE ULTIMATE CRITERION FOR EVALUATION SUCCESS, AND SEVERAL PRECEDENT ATTEMPTS TO PRESCRIBE EVALUATION PROCEDURES TO THIS GOAL ARE REVIEWED. THE SOCIAL SERVICES EVALUATOR IS UNDER SPECIAL PRESSURES FOR RELEVANCE AND IS DEALING WITH A MULTIPLICITY OF CLIENTS WHOSE GOALS MAY BE CONTRADICTORY. A CLIENT MAY BE AN EXTERNAL BODY WITH POWER TO IMPOSE SANCTIONS AND COMPEL CONFORMITY WITH ITS DEMANDS. FURTHERMORE, ONGOING PROJECTS WITHIN THE ORGANIZATION MAY BE EVALUATED FOR A HIERARCHY OF INTERNAL CLIENTS WITH DIFFERING PERSPECTIVES. EVALUATIONS MAY BE INTENDED FOR DECISIONMAKING IN THE SHORT RUN OR FOR FURTHERING LONG-RANGE RESEARCH GOALS. IDENTIFYING PRIMARY CLIENTS AND RECOGNIZING THEIR EXPLICIT AND COVERT POLITICAL PURPOSES ARE PREREQUISITES OF SUCCESSFUL EVALUATION. THE ULTIMATE PURPOSE IN TERMS OF ORGANIZATIONAL AND EXTERNAL IMPACT MUST ALSO BE DEFINED. THIS PROPOSED PROCEDURE USES DECISION RELEVANCE AND SITE OF INTENDED IMPACT AS CLASSIFICATIONS FOR THE VARIETY OF EVALUATION DESIGN GOALS. CRITERIA FOR APPROPRIATE DESIGN, EXECUTION, REPORTING, AND IMPACT WILL DIFFER ACCORDINGLY. THUS, A NUMBER OF DIFFERENT EVALUATION TYPES CAN BE OUTLINED ACCORDING TO PURPOSE, AND THEY CAN BE JUDGED FOR THEIR IMPACT. REFERENCES ARE INCLUDED. (MRK)