This is something I’ve been thinking a lot about lately as we’ve been working on and refining Nincha for the press launch.
Essentially, there are 3 ideologies of UI design:
Set an invisible standard
Take a risk
The key is figuring out where and when to use the right ideology.
For example, with something like your site’s header, you want to set an ‘invisible’ standard. By that, I mean you want to make something that isn’t too out there or copied from someone else, but something that feels natural, real, simple, definitive, and in one word, ‘invisible.’ Why? Because the header of your site is the element that will be seen and used more than any other, and it’s not something you can risk to be different (because they’ll get tired of it) or take/copy from somewhere else (because they’ll notice.) It should essentially standardize the look of your website, and sum it up in a 1024x30 pixel block.
On the other hand, you can’t build something great unless you innovate, and that involves taking risks in UI terms. This means doing something totally differently, changing user behavior from the norm, and succeeding in doing so. You have to carefully choose the right place to do this — a place you genuinely think has room for improvement — because you don’t want to seem like you’re reinventing the wheel or doing something for the sake of doing it. The success rate with this ideology is low, but when it works, it really works. My best example to date (and perhaps the only one) would be the Memiary sign up/login system, and I think it’s worked in my favour because it (a) reduced thousands of unnecessary clicks and (b) introduced users to Memiary’s different design philosophy — making them want to try it further — with merely the first thing they see on the site.
But of course, if you try to reinvent everything and innovate everywhere, you’re going to fail. Even Apple didn’t reinvent the number pad on the iPhone, because they knew that what was out there was already the most user-friendly, standardised option, and they couldn’t change it even if they wanted to. Same goes with the menu bar in OS X (a contrast to the Dock), and it also explains why Firefox looks and works very similarly in Windows, Mac, and Linux. You have to follow conventions and structures where they seem fit, and weigh the user’s familiarity with that norm vs. if it’s worth fixing what they find annoying about it by reinventing it. If it isn’t, go with the standard. If it’s something minor, fix it (the chances are, it’s already perfected itself over the months/years, and that’s how it became the norm.)
In the end, I think a balanced UI with a focus on end-user experience works best. Apple has mastered this. They’ve innovated where they’ve wanted to (among many examples: scrolling sideways from the home screen, or upwards in the iPod screen, to search), followed conventions where they’ve had to (tabs, number pad, keyboard layout, the way many of the default apps work), and set invisible standards where they’ve needed to (slide to unlock, scrolling, keyboard).