Omniture Articles

Tests Bring 267% Lift In ROI For Restoration Hardware

Web executives at Restoration Hardware were committed to improving usability on their site — but even more they wanted to know the truth about their customers. Once they saw the concrete evidence, they were committed to making the changes that were proven to work.

Using Offermatica, they put the usability study through its paces by testing:

  • placement of different elements on product detail pages
  • placement of key gift items during the holiday season
  • cross-selling during the checkout process

The results blew expectations out of the water: The test paid for itself within the first week and returned a 267% ROI over the duration of the test.

"The impact of our usability study was greater because we were able to test the results," says Artemis Patrick, Web Merchandise Manager for Restoration Hardware. But beyond an exciting increase in revenue, the Restoration Hardware team gained an insightful view of testing processes.

Here are five important lessons Artemis shared — lessons that are all the more interesting because they can be applied across industries:

Lesson #1. Use what you learn from the first test and dig deeper

The first test boosted revenue per visitor by 5.28% over the default. The second test boosted it 5.5% over the default.

"One of the things that's great is that you can take your winning recipe and fine tune it," Artemis says. "Your winning recipes have a bigger impact because your knowledge is based on previous tests."

Lesson #2. What other ecommerce sites do isn't always right

Because Restoration Hardware is a three-channel retailer (brick-and-mortar, catalog, online), Artemis says, "We have constant input from people beyond the web team saying what they want to see online. People look at other sites and assume that's the way to go."

By testing the elements in question, the web team learned that some of those things worked and some didn't. "We don't want to do simple changes to the site blindly, because what others are doing isn't always the right thing," says Artemis.

And it's not a matter of being right or wrong or improving an internal image, she says. "Half the time, I was proven wrong. It's about taking all the ideas and seeing all the data."

She adds, "I don't want to come back to an executive and say, 'I don't think that will work.' I want to show if it will work or not."

Lesson #3. What do you value in a visit?

"Just because a visitor seems interested in something and clicks on it doesn't mean it will necessarily increase conversions."

For example, offering retail locations online might increase page views but decrease conversions. "That's not necessarily a bad thing," Artemis says. The point is that you need to know what your goal is, and by what metric you'll be measuring that goal, before you begin.

"Revenue per visitor, conversion rates, those are both great metrics, but understand that you can measure other things, including click through."

Lesson #4. Look beyond the obvious

"It was an eye opener that you can get improvements from testing the smallest details, such as placement of copy information. Look beyond just the home page, beyond testing one image against another, to dig into the customer experience from beginning to end, and know that you can test those things, as well."

Lesson #5. Don't worry about structuring the test until after you brainstorm

Restoration Hardware's brainstorming process was exciting. "We got everyone from marketing, creative, the other groups, all together in the same room to throw out ideas."

With all those ideas on the table, the process could have gotten nerve-wracking. "You want to be sure each test is impactful. Sometimes they're not, but you want to go in with the expectation that there will be great impact."

By looking at each idea in terms of what they wanted to accomplish, they were able to design useful tests.

"The consulting value of Offermatica was even greater than the tech support," says Artemis. "The team really helped us sort through the data and make recommendations on how each test should work."