Obtaining an unduplicated count of persons receiving homeless services is one of the primarly goals of the Homeless Management Information System (HMIS) initiative; however, accurately translating client data collected in an HMIS into an unduplicated count is not simple. This paper proposes a method of unduplication that addresses data incompeleteness and inaccuracies and the need to mask client identity in certain circumstances. While unduplicating on the basis of personal identifiers allows for the greatest level of manual intervention by system administrators and potential accuracy, that approach is not always feasible given data quality and privacy concerns. (abstract from the document)