The script you posted should work assuming the values are within acceptable values for each datepart. As much as I hate to say this I think you may have to use a while loop here and convert these one at a time to find the culprit then. Stick all the parsed values into a temp table with an identity and an extra datetime column(with all NULL). Then you can use a while loop to go through the temp table rbar and update the current row's datetime column with your parsing logic. You will find the invalid data in there somewhere.