Maybe I'm just confused but it appears to me the scenario you suggest requires that the table be populated first. Otherwise I wouldn't be able to do the check. So ultimately I'd be running the append twice; once to populate the table (with the Unique index) then checking aginst the first table when appending to the second. It solves the issue of eliminating the duplicates but is basically the same as Martin's suggestion which would require me to run the append process twice, right? I still don't see how this would work in one pass. Am I missing something ?

This application will be replacing an old mainframe program, which by the way takes over 8 hours to run the same process, so from an operations standpoint it's still no contest as the time would be cut by 3/4 even if it takes a couple hours. It just seems to me that I should be able to do this faster (and in one pass).

Thanks for your help.