THE ANSWER TO THE QUESTION OF WHICH TREATMENT APPROACH IS MOST EFFECTIVE WITH DIFFERENT KINDS OF CLIENTS CAN YIELD REAL BENEFITS TO CLIENT SERVICE. A STUDY OF A NONPROFIT AGENCY OPERATING SIX GROUP HOMES FOR ADOLESCENTS WAS USED TO HIGHLIGHT THE IMPACT OF PROGRAM EVALUATION. A HIGH RATE OF CLIENT TURNOVER WAS CAUSING BOTH CLINICAL AND FINANCIAL PROBLEMS. THE STUDY SOUGHT TO DETERMINE WHICH TYPES OF CLIENTS DID WELL AND WHICH TYPES OF CLIENTS DID POORLY IN THE PROGRAM. DISCUSSION CENTERS ON THE DIFFICULTY OF DECIDING WHAT TO MEASURE, WITH COMMENTS ON SELECTING RELIABLE MEASURES, THEIR TURN-AROUND TIME, AND THEIR INFORMATIONAL VALUE. IN THE STUDY OF CLIENT TURNOVER, THE OUTCOME MEASURE SELECTED WAS A TWO-VALUED SCALE DESCRIBING WHETHER A CLIENT'S DISCHARGE FROM THE PROGRAM WAS CONSIDERED TO BE A SUCCESS OR FAILURE. THE STUDY WAS BASED ON 110 YOUTH SERVED BY THE PROGRAM OVER A 3-YEAR PERIOD, FROM 1972 TO 1974. RESULTS INDICATE THAT 29 PERCENT OF ALL CLIENTS LEAVING THE PROGRAM WERE DISCHARGED SUCCESSFULLY, BUT YOUTH FROM SMALLER FAMILIES WERE TWICE AS LIKELY TO RECEIVE SUCCESSFUL DISCHARGES THAN THOSE FROM FAMILIES WITH FIVE OR MORE SIBLINGS. DESPITE A NONEXPERIMENTAL DESIGN AND AN INTERNALLY GENERATED MEASURE OF OUTCOME SUCCESS, THE STUDY WAS USEFUL IN HELPING THE AGENCY TO EXAMINE AND IMPROVE ITS SERVICES TO CLIENTS. THE KEY TO SUCCESS IN PROGRAM EVALUATION MAY LIE IN THE USE OF MULTIPLE METHODS TO CORROBORATE FINDINGS TO COMPENSATE FOR LIMITED CONTROLS. ONE DIAGRAM AND REFERENCES ARE PROVIDED. (PRG)