Best-Guess Design
Any web site redesign will deliver you the designer's best guess solution, nothing more. Even if you carry out extensive client interviews and usability testing, the choices that are made along the way are all guesswork!
They may be educated guesses, backed up with personal experiences, case studies, and logic, but when it comes down to "How should this navigation be arranged?" or "What's the right style for this callout?", all we can do is give our best guess.
Now I'm not saying there's anything at all wrong with that. In fact, it's called Design, and it's a great process, which has given us a world full of amazing things. The art of creative expression is absolutely essential. You can run a usability test to find out which of 10 layouts people find the easiest to use, but unless you think creatively, you may never find the magic 11th layout that will blow the others out of the water!
But creative guesswork also has its limitations. Let's say you get your web site redesigned. How many changes will be made, to colour, layout, size, position, navigation, copy, style etc.? I bet hundreds or thousands of things will be different.
Creative Creationism
Now, if your new web site has a 5% better success rate (however you measure it: time spent, leads generated, sales made), how can you figure out the way to increase that success rate further? Which of the 1000 changes have helped increase success most? Are there any that have been counter-productive? Have some changes cancelled out other ones?
The fact is, you can't know. Because what you've got is the result of a mass of creative thinking, all delivered at once. That's the web design equivalent of creationism - everything arrives already designed. Take it or leave it.
An Evolutionary Alternative
An alternative approach is to make your design/content changes one by one, in a controlled way that lets you isolate and measure the impact of each change.
We'd look at a design and imagine what creative alternatives we guess might work better, using our knowledge and experience to inform our guesses. (Then we always run our suggested changes past the client for their approval or feedback.)
The key to the process is to make only a few distinct changes to a design in each cycle of the "test & measure" process. (Google Webmaster Tools offers a really powerful way to run split A/B testing that works on most sites, working in conjunction with Google Analytics to provide usable results.)
When you can see the impact of using callout 'B' as opposed to callout 'A', a world of possibility and power starts to open up.
If B works better than A, you then ask why. What is it, all other things being equal, about design solution B that means 10% more visitors chose to click through to the offer? Was it the shorter copy, the bolder text, or the colour difference?
You see, like usability testing, split testing like this can give you data, but not necessarily answers. And it's not creative in itself either, so this is where you need to fire up the old brain!
Step Two is to hypothesise on the reasons behind any significant difference in performance. (OK, "hypothesise: is a fancy word to say we're still guessing at this point, but we have a clue, because we have metrics.)
Once we've done best guessing as to why B beat A, Step Three is to test our hypotheses. (Sometimes, you'll just be happy that B works better, and move on to another area.) To test whether our reasons are probably true, we might do another test in the following cycle, but this time testing option B against a "super-B, i.e. something that's got more of B than B had. If super-B also performs well, it suggests that the B-ness factors are likely to be positive.
What this all means
We think that testing & measuring is usually the best way to invest in a web site. By following this process, our clients can be sure their sites work better every week/month, in ways they can measure and put a value on.
It also gives a great sense of control, at the same time as minimising risk. Clients can manage their budgets much better, knowing that each small payment will give them both success and useful knowledge. If the results show a great return on investment, clients can opt to accelerate the process by doing more T&M cycles. Also, if the improvements slow down, the client can save money by slowing down or halting the process.
When is testing & measuring not advisable?
Some (many) web sites are clearly too badly designed or flawed to be worth improving through a gradual process (ever heard the phrase "you can't polish a turd"?). In these scenarios, it's more cost-effective to carry out a full redesign in the hands of an expert designer, who has a track record in delivering much better best guesses.
Even after you've redesigned, you still have a best guess, so it's time to get scientific and start honing the messaging and design to make sure your site is moving steadily towards optimal performance.
I would recommend anyone who has a web site to pay a similar amount of attention to testing & measuring your on-site flow as you do on search engine optimisation today. There's no point pouring more water into a leaky bucket. Much better to spend your time first fixing the leaks, then pour in the water.
