More on the value of tweaking from Seth Godin. This is a good post. As with most things, the ideal he speaks about is right on but the reality of these tweaks (and measuring them) CAN be much more complex. At Amazon, sometimes things that were "obviously" better would take months to come out because to instrument all of the metrics that needed to be captured to provie it was way more work than the actual code being altered.

Quoted: The opportunity, as the web becomes more sophisiticated and CSS gets implemented more often, is to figure out how to tweak a page while it's running and get 2% better response from that page. 2% isn't a lot--until you multiply it by a million page views.