To me its all about knowing your data. This is a great method to use some of the time, but it will cost you in performance in some cases. At our shop we have a very large database (which happens to be on Oracle), but when we are trying to perform an extract to a data mart (SQL Server) we generally do not want to use an index anyway because the volume is so large that the indexes are many times less efficient than a table scan. In this environment the technique described in this article would be good. Thanks