Multivariable Web Testing Versus Traditional A/B(Cs)
Most marketers have dabbled in testing versions of their Web sites in hopes of increasing conversion rates, but the process can be time-consuming and many companies complain it doesn't work. Mark Wachen, CEO of optimost, explained the benefits of multivariable testing at a breakout session on lead generation at WebSideStory's Active!Insights user conference in New York on Wednesday.
"One of the common misconceptions with wanting to improve Web pages is [that] there's just one decision--go on or not. [Everyone] buys adds on Google to get the horse to water, but how can you get them to drink?" Wachen said. "It's important every part of your page is optimized to increase the likelihood your visitor is likely to continue on."
A/B testing is the traditional process. A company creates two versions of a Web page and finds that one performs better. The problem is, that organization does not know why it did better, because it can't drill down to see that some changes had a positive impact, some negative, or perhaps even canceled each other out. Other companies only test one variable at a time. That's a better approach, Wachen said, but it takes a long time, and "you're not taking full advantage of all the information your visitors are providing." Also, it's not very robust. One image might prove better than another, but if some other variables change, the results may not be the same. "The road to low conversion rates is paved with good intentions," Wachen said. "No one will make a decision about a page based on a bad idea, but you might be missing certain things by making grand-scale changes."
Multivariable testing is taking a page and breaking it into component parts, asking what things a company wants to test. Anything can be a variable, Wachen said. The idea is first to identify variables and then go to each one and identify different values to test for each one. "That could be 15 variables and 80 values that lead to potentially millions of possible permutations," Wachen said. "But you can test a much smaller subset that's representative of the whole to optimize performance. The important issue isn't finding which page converts faster, but whether you're introducing factors that make a difference."
The amount of copy on a page can be overwhelming and the order in which it's presented can matter. Simple headline construction, like turning a statement into a question or bolding certain words can help, and testimonials can work wonders, Wachen said.
He focused particular attention on images and multimedia, which he said are the least predictable element of multivariable testing. "It's not always obvious what image is going to work, but what's always true is to ask, 'What's the point? Why am I using this image?' Is it really conveying the message you're trying to convey? It's important to test continuously," he said. Multimedia technology can be ineffectual if a company forces someone who is not tech savvy to use it; audio can be embarrassing if it's too loud.
"Nearly in every case, multivariable testing is the way to go," Wachen said. "There's tremendous benefit. Be adventurous with testing. There's limited downside and unlimited upside. If something isn't working, you can weed it out right away. There's always room to improve. Until you get 100 percent conversion rates, you can always do better."
Happy Holiday Marketing
"You don't want to wait until next year to see what you should be changing."
Sewing Up Online Offers
Test to understand what compels customers to convert.
Email Delivers ROI
Upped conversion and click-through rates mark that despite declining open numbers the channel remains popular.