I respectfully disagree that incremental loads are faster than the truncate method. The problem with an incremental load is the amount of resources it uses on the SSIS side to manage joins and lookups. The need for SSIS to load the dataset into memory before it can perform a comparison is great. In addition to the resources required to pull up the data in SQL, the amount of resources needed by the SSIS machine to manage the comparisons may even be greater. Part of this is the need to ensure string comparisons are done with trimmed fields, re-sorting, and data conversions that are performed to ensure the comparisons are accurate. I've yet to see a comment somewhere where the accuracy of the comparisons are put to question but they do occur if these are not taken to consideration. This is due also to the fact that the collation of the SQL server is different from the collation of the Windows Server machine that runs the SSIS package. From my experience, we have processed, logged, validated, and loaded about 60 million records in 2.5 hours using the truncated method (with drop/re-create specific indexes), as opposed to the incremental that takes about 6-8 hours.