You need to make sure you gather system statistics with both NOWORKLOAD and then with a representative workload, otherwise you can get some screwy behavior. As to your question about making a difference, YES they can definitely make difference as the COSTS (cost based optimizer) uses several of these values to come up with the cost calculations of any execution plan. See below URL for a good write up about it. http://karenmorton.blogspot.com/2008/06/optimizer-noworkload-statistics.html Chris -----Original Message----- From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Andrew Kerber Sent: Saturday, August 17, 2013 11:36 AM To: tony@xxxxxxxxxxxxxxxxxxxx Cc: ORACLE-L Subject: Re: System statistics Valid points. I have gathered system statistics on several different systems. They have yet to make any significant different in performance, good or bad. Has anyone seen a case where the system stats do make a significant difference? On Sat, Aug 17, 2013 at 11:31 AM, Tony Hasler <tony@xxxxxxxxxxxxxxxxxxxx>wrote: -- //www.freelists.org/webpage/oracle-l