How Google Website Optimizer Encourages Visitors to Act

By daniel - Posted on 09 October 2009

I learned a lot about Google Website Optimizer in a presentation at Net Tuesday Vancouver earlier this week. I also have a write-up on Adsense from the same event, if you're interested.

VKI Studios' John Hossack was the speaker. His presentation focused on the "post-click" portion of online advertising. It starts from when a user clicks on an advertisement and ends up in your website.

The goal of this process is to get the user to perform some action, like purchasing your product, making a donation, or telling their friends about your product or service. There are many different ways you can encourage a visitor to take action. Different wording, button styles, product images, and layouts all have an effect.

So how can Google help? One way to decide which button style makes more users want to click it is to use your experience or intuition. Another way is to test, test, test. And that's where Website Optimizer comes in.

What Website Optimizer Does

Website Optimizer a free service from Google that allows you to present different versions of your website to users. Each time somebody visits your website they'll see one of your test cases. If you're testing button styling, one test case might be a yellow button and another case a red button.

Web Optimizer records which version of the website was displayed and then records whether or not the user clicked the button. This button click is your "success event" and should lead to one of the desired actions mentioned above.

Once you get enough of these "success events" you can look at the numbers to determine which button was more likely to be clicked. By testing different improvements all over your website, you can make it more effective at getting visitors to perform the desired action.

Notes on Website Optimizer

There's a caveat to all this. Your site needs to have visitors coming in and triggering "success events" fairly regularly for this to be useful. If only ten people click the button in a month then you don't have very much data to measure. John recommends having at least 100 success events per test case per month.

"Test shouts, not whispers."

John also points out that you should be testing big changes. He likes to use the six-foot rule: you should be testing changes that are visible from six feet away from the monitor.

I have to say I learned a lot from both speakers at Net Tuesday. Many thanks to the speakers! Slides for both presentations are now available to Net Tuesday Meetup members.