Designing by the Numbers

Online retail flew high in last year’s last quarter, posting a double-digit increase over a year ago. But with the last bit of tinsel packed away, it’s back to the reality of retail as a 365-day-a-year business. Absent holiday spikes, online retailers must still look daily for sales. That means looking beyond the broad impact of a few must-have gift items or killer seasonal promotions to smaller gains to be found throughout the site and across a wider spectrum of the customer experience.

To find those opportunities, smart web retailers are using analytic tools that yield data so immediately illustrative of customer behavior that some in the industry now talk of web page design as a design-by-numbers game.

“The low-hanging fruit has been picked,” says Doug Greene, chief technology officer and chief marketing officer for Overstock.com Inc. “The easy wins have been gotten. What you want to do at this point is get that .05% increase in conversion by trying this or that with the web site. You have to be scientific about it, with a control group to quantify the effects of changes to the site. And you can’t do that without these kinds of testing tools.”

It’s all about usability

Jupiter Research Inc. analyst Matthew Berk notes a surge of interest in measuring the performance of web sites, not only to justify money spent, but also with an eye toward optimizing the site. “It’s all about usability,” Berk says. “Most of the work that sites have done on usability has been myopically focused on the customer—let’s take surveys to find out what they want and who they are. That’s great, but it’s background information. What you really need to know in managing your site is what you want customers to do on the site and how effective the site is in getting them to do it.”

In other words, understanding actual behavior is more useful to site operators than understanding visitor demographics, which after all, say experts, only hint at how people will behave online. Achieving that understanding is the goal of a rapidly growing field of web site analytic tools.

Toward that end, Berk has praise for Overstock’s use of analytics to manage by the numbers. “It’s a complete data-centric approach to managing what they sell and how they sell it,” he says. Overstock had been capturing, testing and measuring the effect of site changes on conversions internally, but boosted its analytics power by implementing Omniture Inc.’s SiteCatalyst analytics product in December.

“The biggest benefit in using the tools is being able to quantify changes we make to the site,” Greene says. “If we change the color of the Buy button, I want to be able to measure that change’s impact on conversion. A company like ours might have 100 campaigns going on any day. Given all the noise, it is impossible to quantify the results of small changes to pages, and whether they help or hurt conversion, unless you have this kind of tool.”

Greene adds that he saves the cost of one or two full-time staffers otherwise needed to generate the reports that Omniture’s hosted application produces. Greene won’t disclose what Overstock pays for the service, but the cost structure is a fixed fee for bandwidth use up to a designated level, with higher volume-dependent fees after that level.

Measuring variables

Web retailers are now using analytics tools to measure everything about the customer’s site experience from loading times to different shopping technologies to the presentation of content and merchandise on pages. And those measurements are informing improvements that show measurable results. San Diego-based flower site Proflowers Inc., for example, recently put analytic data from WebSideStory Inc.’s Enterprise product to work in a page redesign at Proflowers.com.

Using the tool to follow the behavior of site visitors, Proflowers tested response to some 20 elements on its landing page by setting up a series of test pages and directing selected customer groups to them. The tested elements primarily involved the use of real estate on the page, such as number of products or a product’s location.

“We identified variables on these pages that we thought might have an impact on conversion, and we set up a matrix to test the elements against each other,” says Richie Hannah, web strategist at Proflowers. Visitors from selected major engines were routed to the test pages while control visitors were routed to the former landing page. After several rounds of testing response to different elements, and statistical analysis to determine how each element affected conversions, Proflowers put the winning elements on a test “superversion” of an optimal landing page.

Over a period of about six weeks to gather data and test response, the newly optimized landing page containing all the top-performing elements drove a 10% to 15% increase in conversions, says Hannah. The analytic tool tagged visitor response to each of the variables, so Proflowers could see, for example, that version three of the landing page was driving more conversions than version four.

“We’ve had opinions of what we thought would help or hurt conversions, but we’ve seen time and time again that what seems to make logical sense isn’t necessarily true,” Hannah adds. “The only way to really test that and prove it is to have people come to the site and vote with their mice.”

Likewise, Tower Records saw click-throughs on featured offers at TowerRecords.com rise by a healthy 10% to 20% and sales increase proportionally after it redesigned its home page based on site visitor data supplied by Fireclick Inc.’s analytics reporting, says Kevin Ertell, Tower’s senior vice president of online operations

Tower got a few surprises after implementing Fireclick’s Netflame analytics tool and an accompanying product, Site Explorer, in August, Ertell says. The analytics provided aggregate data on where visitors went on the home page—and where they didn’t go. “We had special offers on the page where people weren’t clicking at all, or very little,” he says.

First stop: search

Analytics also showed that while visitors were bypassing the offers, they were flocking to search. 65% of home page visitors went directly to the keyword search function, a greater number than Tower had estimated. “We knew it was high, but we had thought there was more browsing,” Ertell says.

Tower used the data in a redesign. The search box was placed more prominently on the left side of the page to draw even more initial clicks when visitors landed on the home page, and a much larger block of real estate immediately to its right, in the middle of the page, was devoted to promotions. “We wanted to make the promotions bigger and more obvious so you couldn’t possibly miss them, even if you were going to the site to search for a specific item,” Ertell says. Since the change was made in October, Ertell says clicks on the featured promotions have risen from an average of less than 1% to 10% to 15%, with sales on those products increasing correspondingly.

Though he adds that most of the data supplied by the analytics tools could have been pieced together from Tower’s internal monitoring, it was easier and less time-consuming to identify in the reporting delivered via analytics. “The picture was worth a thousand words,” he says.

IRCE 2008 Conference Multi-Media CD-ROM

At cataloger and web retailer Hanover Direct Inc., the company’s six major brands have been using Coremetrics Inc.’s Marketforce analytics for the past three years, and the tool drove much of the redesign of its Domestications brand’s home page last year.

“We used a lot of quantitative data from Coremetrics to help the designers in their decisions on what the pages should look like,” says Jonathan Kapplow, Hanover’s marketing manager. “Domestications is mostly soft goods but it has a very deep product base, so figuring out how to categorize it all so as to require a minimal number of clicks to convert was a real challenge. We used Coremetrics to run a lot of historical analysis to determine what categories should be on the home page, how they should look, what they should drill down to and what items should be placed next to others for cross selling.”

While a single change can produce an impressive double-digit increase, retailers aren’t ignoring the cumulative effect of smaller changes suggested by new analytic data—such as Overstock’s pursuit of a .05% increase in conversions. Site improvements that can tip the scale for even a few customers hanging in the balance over a purchase decision add up over time.

While Proflowers got a significant boost from its optimized superpage, for example, it’s also used analytics to drill down deep into customers’ on-site behavior to change details that benefit conversion. When analytic reports revealed that 96% of visitors were JavaScript-enabled, Proflowers optimized the site with features for that customer segment. Those visitors, for example, can select from a previous recipient list to automatically populate name and address fields without having to click additional buttons. “The site works perfectly well for those without JavaScript, but visitors who have it get an even better user experience,” Hannah says.

Photo site Image-Edit.com likewise fine-tuned its home page based on analytics and realized a small but measurable sales increase as a direct result. After it altered its home page to give some of its lesser-used products bigger play, based on data from NetIQ Corp.’s WebTrends analytics reporting, Image-Edit.com saw interest in those services increase almost immediately.

Restoration of old family snapshots is the most used service on Image-Edit.com. Customers scan the photo, either at home, or at a retail camera store that offers the service, and the file is transmitted digitally online to Image-Edit.com for service. The enhanced image file is sent back digitally to the store for printing and pickup, or to the customer’s home computer, where it can be loaded onto a disk and taken to a photo processor of the customer’s choice.

Though photo restoration is the most popular service, Image-Edit.com also lists 16 other photo editing procedures on the site that weren’t getting as much customer attention. “The question was, if we highlight another service, will that make visitors interested enough to check it out,” says Aaron Daru, vice president of operations.

Image-Edit.com put that to the test in June with its Hollywood glamorization service, an image-enhancing process that makes Hollywood-style glamour shots out of basic photos. It opened up a fourth of its home page real estate to highlight a rotating roster of services, with the glamorization service rotated in frequently. Analytic reporting showed that clicks on the service rose to 4% of visitors from less than 1% almost immediately, says Daru, with sales of the service increasing at the same rate.

New level of science

When Hanover questioned the use of a single tiny piece of real estate on its brands’ home pages, Coremetrics’ analytics tool supplied a definitive answer. Like many retail sites, Hanovers’ home pages have a tell-a-friend feature, a small box located on the left side just under the navigational bar. “We realized that this was pretty expensive real estate and we wondered, does this really work,” says Kapplow. “Unless we had an analytics product, we would never have known how many people use it. We probably would have pulled it from the site, but analytics told us there was a nice demand driven from it.”

The data to be gleaned from analytics tools are bringing a new level of science to site design and management decisions that until lately have depended on conventional wisdom, aesthetics, or gut instinct. It’s a speedier and increasingly more precise take on how marketing, merchandising and customer service decisions are supported in the offline world.

“Certainly, when you change anything on the web site you have to back up what you do,” says Kapplow. “In our catalog business, people do page analysis by the square inch to understand the effects of the tiniest change. The web is becoming more like that in that every little thing you do on the page has a dollar ramification attached to it.”

Information, not action

While some retailers are just dipping their toes into web site analytics, others are already looking to the future and wishing for even more sophisticated tools. “I’d love to be able to compare time periods, like comparing our Best of 2002 promotion to our Best of 2003 promotion next year,” says Tower’s Ertell. “We can’t compare pathway analysis for two periods like that in a form that’s easy to view. And I’d like to see something that could do a hardcore regression analysis to tell me what factors really make a difference. There are data mining tools out there, but they are very expensive.”

Some retailers caution that analytics results are only information, not a plan of action. While current analytics reporting puts more—and more user-friendly—data into the hands of marketers, it’s still up to marketers to prioritize their response and take action. “Once you have information, it’s not immediately clear what you can do to increase those numbers,” Daru says.

Hanover Direct’s Kapplow points out that analytics data still need to be linked to a recommendation to have any effect. “It would be great if you could somehow automate that—and then have it execute that recommendation on the site automatically,” he says. “It would be phenomenal if you had a system that was always using data results to populate the various places on the site where we show product, based on what is selling.”

Decisions vs. reports

That scenario is a ways off—at least, at any price that would make it practical for widespread use. But Eliot Jacobsen, COO of Omniture, says the company will begin building some capacity to help guide marketers’ use of analytics data into future releases of SiteCatalyst scheduled to start this year.

“The three major drivers of optimal web site activity are customer acquisition, conversion, and retention,” he says. “We’ll begin to build out the interface and how to approach your job as a web analyst using our tools to align a site along those three major drivers. You’ll see forward-thinking providers in the analytics industry start to guide the process of site optimization instead of simply throwing information at it like a scatterfield of darts.”

As the marketplace for analytics develops, vendors will increasingly take such steps to position themselves as providers not of reports, but of results, says Jupiter’s Berk. “They’ll define a set of business events, such as conversion, registration, add to cart, and then figure out how to optimize against them,” he says. “The hope is that the industry will start to build tools that don’t just run reports, but help users make decisions.”

mary@verticalwebmedia.com

Click here for the Guide to Web Analytics Solutions