When it comes to analytics, Brad Wolensky, Director of Ecommerce for Orvis, thinks 80% of companies that use an analytics package are actually using only 20% of the data they get. "We're one of the more aggressive users of the tool and we don't even think we do it as much as could be done," he says.

That's because there's so much information available, and looking at the data does no good unless you do something meaningful with it, states Wolensky. "That's easily said, but everyone has way too much to do and we know people don't just increase headcount." 

Faced with this problem last spring, Wolensky started making noise internally, telling the president that he could increase sales per visit if he could bring on a Web merchant/analyst to look at data on a daily basis and then make changes to the site. Orvis' president said, "Sounds great, prove it."

Wolensky borrowed a senior merchant/marketing team member to run the proof-of-concept for six weeks, from June 1 through July 30, 2005. He was asked to increase sales per visit (SPV) in five categories (men's, women's, gift, home, pets) by 10% -- a lofty goal.

By the end of the experiment, the goal had not been reached. However, sales per visit had increased 2% to 5%, depending on the category. And that was during the slowest sales season of the year. 

Wolensky got the green light to hire three people (two internally, one external) to run the program. "Since then, the results are conclusive that it is a powerful thing to do. They are paying for themselves multiple times over, there's no question about the value of their cost.”

Here are the five tactics he used to make the program a success.

     -> Tactic #1. Don't hamstring the effort with design rules

Wolensky purposely didn’t blinker brainstorming with creative rules. In fact, he says, the president of the company told the team it was welcome to come up with changes for the home page and to try various concepts. All he asked was that they didn't change the green Orvis logo (really).

     -> Tactic #2. Look for quick, actionable ideas

To improve SPV in a six-week time period, Wolensky knew the Web merchant/analyst would have to move quickly. He hired a freelance art director to be ready to react to the suggested changes on a daily basis (products being rotated in/out, etc.).

Plus, the team decided against running classic A/B tests for the short-term.  Wolensky absolutely planned to implement A/B later on, but it wouldn't make enough difference in the very short time frame he had to prove concept.   Instead of A/B testing, his team rotated items in and out on the site every day or two to see where they could get lifts in sales per visitor (SPV).

     -> Tactic #3. Laser focus on the home page

Wolensky knew the home page would offer important insights.

   a. Monitor new items and add or remove within a day

By looking at analytics and making changes to the site, the team learned that items on the home page with a 10% or higher conversion should be considered super-winners and shouldn't be removed. Items on the home page with conversions lower than 3% should be immediately taken off.

"For a fairly well-trafficked site, you should have enough data to evaluate how [an item] is doing within 24 to 48 hours, and I don't think people do that," Wolensky says. "They let it sit for a week or two or a month."

   b. Keep high converting categories represented at all times

Wolensky also learned that certain categories have higher conversion rates but lower traffic, while others have higher traffic and lower conversions. Using analytics, his team found the products with high conversions and high sales per visit and learned that they should be maintained on the home page at all times, either via a product shot or a banner that represents the product. For example, while the current home page touts (via product shots on the main section of the page) men's clothing, women's clothing and fly fishing, both pets and home are each given a banner across the center of the page. 

   c. Look at navigation

The Orvis site has four ways to navigate from the home page: a left nav bar, a top nav bar, the middle banners and search. By watching how customers tended to shop on the site, Wolensky found that the left nav bar was used most, so his team made sure it was as "good as it can get."

During the six-week test, the team added three new links to the taxonomy on the left bar, which now remains relatively static with 14 choices. Once a customer navigates away from the home page, the left nav bar "explodes" to include many more. When a visitor clicks on Men's Clothing, for example, the bar includes 29 choices.

Interestingly, while the team learned that the home page navigation bar should remain relatively solid, the taxonomy on the category pages can change. Rather than being in alphabetical order, the top six or eight best categories are listed at the top, and those change depending on sales.

     -> Tactic #4. Cross-pollination of categories 

Wolensky's team learned that just because a customer is looking at gifts it doesn't mean that they won't be interested in other categories, such as luggage. "If they're on the gifts landing page, we should also have some other products in there to keep them browsing. If you slice it too narrowly, they may bail out from the site."

        -> Tactic #5. Appeal to a variety of visitor personas

Orvis struggles as a business with a diverse product line, from men's fly fishing to women's apparel. The home page must reflect that balance by promoting what's converting without turning off other segments. For example, the current home page (which changes weekly) has three product shots across the middle of the page: the first is a vertical rectangular box with a man wearing fishing gear. It's slightly smaller in width than the next two boxes -- a men's sweatshirt and a woman in a crisply pressed white blouse.

"You need to use your data to say what you should be promoting, but you have to be sensitive to the fact that you can turn off some people," says Wolensky.

One final lesson learned:

"We think we've learned that women seem to like the search function more than others and use the search box more than other segments," says Wolensky.   So he is currently working on a project to improve the search results presentation so it's more visual. "In particular, we feel that for women, where you're talking about apparel we need to give better visual results."

Useful links related to this article:

Coremetrics, the analytics package Orvis uses:

http://coremetrics.com/

Orvis:

http://orvis.com/