So you’ve got a stellar website design with copy that virtually sings. So why is it such a fantastic flop?

“Copywriters and designers are often not the users of a site,” said Vita Janecek, online marketing product owner for Kentico Software. This means that their designs and words, while they may look and sound good, may not appeal to your target audience.

So how do you get it right? It’s about going right to the users and asking them what they want, said Janecek, who spoke at Kentico Connection in Boston recently. Early testing can ensure that you’re hitting the mark.

After all, in the words of an old African proverb that Janecek quoted, “Nobody tests the depth of a river with both feet."

A Toe in the Water

Testing can tell you ahead of time how well your website and email will perform, said Janecek. You can choose different tests depending on your goals.

A/B Testing involves showing two different designs or email to website visitors. Traffic to the site is split to the two different versions of the site so you can track which is more popular.

Multivariate testing enables you to test single elements of a site. This type of testing typically requires more visitors to gauge statistical significance accurately, said Janecek.

The main result you want to look at during testing is your conversion rate, he said. When it comes to A to B tests this means looking at which page or email leads to the most conversions. Testing can look at individual pages, such as landing pages or those that are part of your checkout process to see which ones are most effective, he added.

Reinventing the Wheel

When you start testing, it's not the time for minor tweaks. Make radical changes, said Janecek. Get rid of your current design and message completely and try something new.

“In the beginning of testing is the best time to learn if a new design will help, he said. This might mean changing colors and text, the position of different items and remove anything confusing.  A/B testing will differ slightly if you’re using it to determine how well your emails are hitting the mark. In addition to measuring conversions you can also measure open rates. You can use tests to find the best components, such as including the sender name in the email or changing the subject line.

Send out two versions of the email: one version to 10 percent, the second version to another 10 percent. This will help you determine which version performs better. You can then send that version to the remaining 80 percent.

Before you finish a test, ask whether you’ve had enough visitors to get good results. As a rule of thumb you should strive for approximately 1,000 visitors and at least 100 conversions to get good data, Janecek explained. “Most importantly look for statistical confidence which should be indicated by every good testing tool."

The testing period should be long enough to give you steady results, which typically takes about two weeks. Your process should always begin with clearly defined goals and objectives. Create a list of your web pages and decide where you will start testing. It’s wise to start at the top of your list with the items that are your highest priority.

“Once you have tested the first page, you can begin ongoing testing,” he said. If you’re not getting the results you’re looking for, it’s time for another round of radical changes.

While initial tests may give you great improvements right off the bat — gains from 20 percent to  200 percent — that will tail off over time. But that doesn’t mean that testing should slow down. It should be ongoing to help you fine-tune your site. “Your conversion improvements will be lower, but they will still be there,” said Janecek.