But if you operate a working, established, or otherwise popular website (say at least 2,000 visitors per day), I would never recommend a major visual or mechanical overhaul. It pisses people off. And when that happens, loyal visitors flock to alternatives in mass exodus, as Digg users have done this month.
There are a couple of exceptions to this rule. If your website has a monopoly on information, you can do whatever you want, and readers will keep coming back. And if your website isn’t “working, established, or popular” to begin with, you only stand to gain from a major overhaul, provided it’s done by someone who knows what they’re doing (aka no flash, proper xhtml/css coding, a regular content plan, and most importantly, good usability).
What can you do then to improve or refresh established websites? My advice is to make subtle changes to your design and monitor your visitor’s behavior. If the change has no significant effect, or better, a measurable improvement, keep the change. If the change is off-puting to visitors, revert to the the previous version immediately and re-evaluate both your desire for change and your strategy.
I know this holds true on the few “popular” websites I publish. And if Digg is any indication, I know it holds true for mega websites as well.
May all your redesigns be well-received.