We did this in our office for a client once... First, we normalized the database to USPS standardized text, based on the chart of accepted postal abbreviations: http://www.usps.com/ncsc/lookups/usps_abbreviations.html That alone turned probably 90% of the potential dupes into exact duplicates as for the rest -- if there are multiple apartments/suites/floors on the same address, we suggested that they send that anyways -- there isn't any real way to tell that address isn't a separate entity. could be neighbors or co-workers who should each have gotten their own postcard.