Measure the site performance internally, externally, globally and deal with mistakes intelligently and in a consistent manner.
Internally, within your site network or sub domain, create a slave server to run steady heartbeat pings or page requests against your web server. Monitor the response time to this heartbeat across your 24 hour day. A sample per minute should be sufficient to show performance bottlenecks during the day.
Externally, set up a public access server that runs the same heartbeat page requests against your server from outside of your corporate network just as any local user would attempt. Monitor this response time across your 24 hour day. This will give you the bottlenecks from an outside user's perspective.
Use a global monitoring vendor to get the global picture of the performance of your site across the day. I have used both Keynote and Gomez services to monitor my systems in the past. Both vendors have many global locations and can give you immediate alarms or scheduled reports that monitor the performance of your pages globally. Both vendors are high quality and the choice depends on the economics of your individual monitoring requirements.
Finally, be sure to monitor broken pages, i.e. 200, 302, 404, and have error handling routines to deal with the disappointed users that discover these mistakes.