Very little is currently known about patterns of shelter use by homeless and runaway youths. The goal of this study, conducted in an Israeli drop-in shelter for homeless youth, was to predict a major dimension of service "output" - namely, youngsters' destination at departure, by characteristics of service "input" (residents' background and entry variables), and service "throughput" (the experience during shelter stay). The sample included 399 residents who entered the shelter between January 1995 and May 1997. Data was drawn from entry and exit forms routinely completed by shelter staff for each of the residents. A discriminant analysis performed on nine input and throughput variables derived two functions, which differentiate three destination groups: youngsters who returned home, those who were placed in a group or foster home, and those who left to an unknown or unconventional destination. The discussion focuses on the meaning and implications of these apparent success and failure cases.