Web Analytics Special Report: Base Your Site Changes on Testing
When we read past the headlines, we learn that the company made a few simple changes. It changed its headline, for example, or removed certain graphics or changed the location of the navigation bar. We might try some of these changes ourselves, then wonder why our results are less spectacular.
Here's why: SuperSmart Store marketers did not base those seemingly simple changes on intuition. They probably even found that many of the changes that increased conversions were counterintuitive. So how did they know what to tweak? They came up with various ideas, then tested them.
Testing online has grown quite simple, and direct marketers long have been familiar with the practice. Yet online retailers, while spending time and money designing their e-commerce sites, often neglect to test the elements that go into those sites. Instead, they base the site design and promotions on what they think will work. Intuition has its place, but consumers are the only ones who can really tell a retailer what will increase sales.
Take Jo-Ann Stores Inc., a specialty retailer of fabrics and crafts. The Hudson, OH-based retailer, whose site at www.joann.com targets arts and crafts enthusiasts, wanted to improve conversions and average order value. The company wanted a way to test different creative concepts on different category pages and increase sales simultaneously.
JoAnn.com teamed with Offermatica, a company whose plug-and-play testing software is used to set up tests, track results and change variables on the fly. Through multivariate testing, JoAnn.com tested various elements at the same time. And, because it works with test cells as small as a few hundred, it didn't need to wait for hundreds of thousands of visitors for accurate results.
Plus, multivariate testing ranks each element alone and in relation to each other, so JoAnn.com executives could pinpoint winning "recipes," or groups of elements that work best together, without having to test each combination of elements separately.
Here's what was done:
Sewing machines category page. The retailer's team devised variations on art, featured products and promotions. In particular, they tested free shipping versus "Buy 2 or more White machines and save 10%." The promotion seemed like a no-brainer. "Who needs two sewing machines?" was the thought. But the team was willing to test it.
Furniture category page. JoAnn.com tested variations on featured products, 30 percent off all items, promotions available only to online customers and free shipping on an entire order.
Shopping cart page. JoAnn.com tested various cross- and upsell options.
It turns out that the winning recipe on the sewing machine category page increased average order value 137 percent, revenue per visitor 209 percent and conversions 30 percent. And, counterintuitively, the "Buy 2 and save" offer beat the free shipping offer. Consumers got their friends together, leading JoAnn.com to sell enough machines to outperform single purchases.
In fact, recipes including the "Buy 2 and save" element increased average order value 74 percent and total sales 110 percent.
Surprisingly, the first cross-sell test on the shopping cart page decreased sales. It took JoAnn.com, working with Offermatica, another two rounds of attempts to find a cross-sell recipe that consistently increased average order value. The cross- and upsell tests on the shopping cart page initially didn't perform as hoped.
Offline retailers have made impulse shopping an art form, but online retailers have been less successful at getting buyers to add last-minute items to their shopping carts. JoAnn.com first presented buyers with low-priced cross-sell items, hoping they would buy them on an impulse before checking out. But buyers didn't take the bait. Instead, they wanted higher-priced, generic items like scissors.
JoAnn.com will continue testing as part of its ongoing business plan. There are a million reasons to test. Retailers have tried most of the basic stuff online. Now they need a better equation to increase conversion rates, and it may not be intuitive. Testing takes the risks out of trying counterintuitive ideas.