Longitudinal designs are effective for the evaluation of innovative social programs, but attrition can be a significant problem, particularly with hard-to-follow populations such as persons who are homeless. Tracking strategies for locating participants are essential, but retaining participants requires anticipating and addressing participants' needs at every stage of the research. A strategy that emphasizes appropriate interviewer characteristics, the relationship between interviewer and participant, and participants' "investment" in the research is critical. In other words, evaluation researchers can improve the retention of even hard-to-follow study participants by adapting research design and procedures to be "participant-centred." An example is given of a program evaluation in Ottawa, Ontario, that implemented strategies to adapt to the needs of persons with severe mental illness and a history of homelessness.