When it comes to website design, don't be tempted to blindly follow the "rules." Sometimes the rules won't apply to your site, where your analytics will always inform smart design choices. 

Until quite recently, designing a website was a highly subjective enterprise. In the best-case scenario, companies would hire seasoned design agencies that had, collectively, decades of experience in design, user research and usability testing. Unfortunately, the vast majority of organizations don't have the budgets to hire these rock-star design teams. Instead, they hire designers who, despite their pleasant-looking portfolios, do their best based entirely on their instincts and on subjective feedback from the client.  

The truth about this kind of web design is that it's completely unscientific. Copywriters and designers make hundreds of decisions when producing a single web page, and if they're wrong on enough of them, they deliver you a page that looks good but is full of hidden inefficiencies and opportunity costs. We've all been there before. A designer comes back with some mock-ups of a new page and lays them out on the conference room table, and the marketing managers sit around and say, "I like that one" or "That one is great!"

This is voodoo web design at its finest: Critical design decisions that could improve -- even double -- the effectiveness of a website are filed away in a drawer, never to see the light of day because the highest paid person in the organization (HiPPO) decided he/she liked the page that had a picture of himself/herself welcoming new customers to the site. 

Today there's a better way. Instead of arbitrarily designing a page based on semi-informed opinions, any website owner can now employ web analytics tools to dispassionately identify exactly which web pages matter the most and which are the most problematic (and sometimes, which part of a page is under-performing). What's more, they can use design-experimentation tools to show random visitors different designs of the same page and scientifically measure which design actually was most successful.  

This revolution, from voodoo to data-driven web design, is already well in place at some of the most forward-thinking organizations. Top retailers, professional services firms, industrial companies, and even political campaigns have quickly embraced advanced web analytics and content experimentation tools to dramatically improve their websites. It's not an exaggeration to conclude that websites that effectively leverage these tools are three to four times more successful than their counterparts (in terms of sales, visit length, repeat visitors, or any other measurement you choose).  

As the web evolves, two kinds of websites are emerging. Of the two, those that embrace and leverage cutting-edge analytics and experimentation tools attract more repeat visitors, enjoy higher ROI on their advertising dollars, and outperform their competition by leaps and bounds. Those that continue to fly in the dark are falling behind -- often unbeknownst to their owners (how many prospective customers will bother to tell you they tried to use your site but got frustrated by poor design and went to your competition?). We're at a crossroads, and for the good of the internet, we hope every site will evolve, improve and become a high-performing asset for its organization.  

As an example, consider StairSupplies from Indiana. They track their key metrics religiously with Google Analytics, and they decided to see if they could improve their homepage, since that's where most orders started, and where many prospective orders ended. They tested a few alternatives using Google Website Optimizer and made a couple of small changes that yielded big improvements: Their online conversion rate increased by 144 percent, and average order size increased by 18 percent.

Can you guess which of the two screenshots below was the original homepage, and which was the winning redesign? 

It turns out the first version, with the added image and the visible drop-down menu on the left, did worse! Best practices would dictate that an image does better; this example shows that even "tried-and-true" design principles should be verified with real data. Had StairSupplies not tracked these metrics or run a scientific experiment, they might have implemented a design that didn't help sales, or that actually hurt them.

If you're in the voodoo camp but now ready to reap the gains of data-driven design, here are some steps you can take to make the transition: 

• Use web analytics to track activity on your website and get a better sense of what people value, which pages get the most traffic, and where visitors come from.
• Determine the outcomes you care most about on your site and track them constantly with web analytics (e.g., sign ups, donations, sales, leads).
• Test new web page designs and content changes to maximize the conversion potential of every visit.

If you've never tested your website before, here are some good things to consider changing:    

• Images: Try bigger, smaller, color, black & white, with models, no models, etc. 
• Headlines: Try questions, shorter ones, emphasizing different points. 
• Call to action: Try different button designs and copy. 
• Layout: Try three columns, two columns, one long scrolling column.

And here are some other tips for first-timers:  

• Test a few big changes, not several small ones. You should have about 100 conversions for every page variation you test (e.g., if you get 300 conversions a month, test up to three page variations total, including your current page).
• Let your test run for at least two weeks. Even if you see results early, it's best not to jump to conclusions and to make sure the data has an opportunity to normalize. When we test internally, we usually wait about a month.
• Make testing an ongoing process. Once you find a winner, keep trying to beat it. If you don't find a winner in your first few tests, remember that no winner is sometimes as helpful as finding one, since you learn what doesn't work and you protect yourself from making changes that could have harmed your performance permanently.

Perhaps the most exciting part of this revolution in web design is that the entire web is improving. Only a few years ago, it was full of sites driven by voodoo decisions. Navigation and reading were harder, and most pages weren't very compelling. Today, more and more websites are intuitive, easy to use, and effortless in accomplishing our goals -- be they reading a news article, buying a book, or hiring a realtor. Have no doubt: almost all of the best designed customer experiences were the result of web analytics and website testing.

Tom Leung is senior product manager for Google Website Optimizer.