If you use the Internet Wayback Machine to view some of the more popular sites on the Internet, like perhaps Google or Amazon, you'll notice an interesting phenomenon. They don't look all that different in 1998 or 1999 than they do at present. But that seems somewhat odd, when you consider that most Web designers spend most of their time redesigning websites for people. But the fact of the matter is, when it comes to retaining customers, redesigns don't work.
Redesigns Don't Work
Redesigns rely on the principle that whatever is on the site at present is so absolutely horrible that no customer is finding anything on it. In fact, if a website is getting any pageviews at all, chances are good that a major redesign will at best generate a lot of complaint email from customers or at worst drop those pageviews significantly.
People don't like change. When a website they are comfortable with changes, it means that the reader has to go and re-read the entire site to find what they were used to finding easily and without trouble. In fact, if you watch someone going to a website she visits regularly, they will actually do things like move their mouse in preparation of the page loading. For example, if the main page has links to comics in the upper right, and that's what she reads first, the reader will load the main page, and then move her mouse up to the upper right, even before the page has started loading. If you redesign the page and move the comics link to the middle or the lower left or (heaven forbid!) remove it completely from the page, you'll both annoy and alienate those readers who are comfortable where it is.
Why Do Companies Redesign?
I think that the reason that many sites are redesigned is because the designers change so often. Every time a new design team is brought on board, that team wants to both show how valuable they are and make a splash on the site. The easiest way to do that is to completely change the look and feel of the website.
But a massive and completely overhaul of a website is very hard on the customers.
Instead, why not imitate what some of the best sites do, specifically, iterative design. Iterative design is the idea that rather than changing everything on a site all at once, instead you make small changes more regularly. For example, rather than changing the entire website, you might make a small change to the elements in your navigation bar or possibly even in just the color of your navigation bar.
But it doesn't stop there. Once you've made the change, you need to track how your customers respond to it. Some changes will be helpful to your customers, others may cause problems. With iterative design, it's possible to make a change, realize that it doesn't work, and back out. All without many of your customers even noticing that you'd made a change. With a major redesign, if there's a problem, you often have to completely change back, which can be just as jarring for customers as the initial redesign.
Some people argue that iterative design is just "lazy" design. That if you can't be bothered to plan out your design in advance, you'll do an iterative design and just keep tweaking it. The irony, from my point of view, is that doing those constant tweaks actually takes as much, if not more, time to do as a massive redesign. The difference is that it's handled by a team who knows and cares about the fate of the website rather than by a design firm who just wants to make a big splash and get out.
Iterative design is not an excuse to just use your customers as guinea pigs, but it does allow you to try things, and pull things out that don't work. The reality is, there are some things that work great in the lab, but when your full customer-base sees them, fail miserably. While there are other things that seem to be either minor or make no difference in testing that are a huge hit with your customers. With iterative design you can minimize the first while taking advantage of the second.