Power Up

Guide to E-Commerce Technology

When summer’s over, colder water sends kayaking enthusiasts in search of protective neoprene clothing to extend the season. One place they can find it online is at NRSweb.com, the e-commerce site of paddle sports cataloger NRS. A search under "neoprene" brings up 277 items, topped by a list of more narrowly-defined, related search terms to use in refining results, if needed. But chances are good that the first item on the results list—NRS Rodeo neoprene shorts—is just what the searcher is looking for because it is the most-searched product in the category. One click and the shorts are in the cart, with a drop-down menu then offered to capture details such as size.

With the typing in of a single search term, the visitor has found what he wanted. But if you think that with this transaction, the site search function on NRSweb.com has completed its job, think again. Today, site search technology can do much more than simply help shoppers find a product.

Continuous, closed loop

Now, analytic tools attached to site search, either through partnerships between site search and web analytics providers, or created by site search providers directly, can inform site merchandising and marketing based on customers’ ongoing interaction with the search box, in a kind of continuous closed-loop model. Marketers are tapping customers’ search behavior on their sites to rank search results, buy online advertising, identify and respond to trends, create on-site promotions and launch e-mail campaigns. In some cases, the combination of site search results, analytics and business rules even adjusts site merchandising and marketing, as those data are compiled, automatically.

At NRSweb.com, where site search is provided by SLI Systems Inc., each customer search on the site is recorded in several reports, some of which e-commerce manager Keli Keach sees daily. One report identifies the top 10 search terms on the site that are producing the fewest click-throughs. For example, if a shopper searches for an ice chest under the term "cooler"—a much broader category than just ice chests in the site’s product catalog—that search might not produce a lot of clicks, though it’s a term that’s searched frequently. "Ice chest," meanwhile, does well enough on its own with clicks and conversions, information Keach gets from another report.

Blending data

She blends data from the reports to track down copy listings in need of adjustment; in this case, making sure the search results listing for "cooler" incorporate the words "ice chest," where appropriate. By adjusting the copy in the site search listings, Keach says she can boot a site search term off the list of poorly-performing terms in as little as a few weeks. When it disappears from that list, but remains a searched term as shown in other reports, it’s because searchers are finally clicking on the listing and finding what they want. That adds up to more conversions—but site search’s contribution to the top line doesn’t end there.

Keach uses a daily report on the 50 top-searched terms on the site to spot trends and adjust marketing. "I can use the words and phrases they are searching for, apply that to the copy on different products, and start highlighting those products on our home page or in our e-newsletter," she says. "I can use that information to change our advertising—by buying a banner, for instance, somewhere on those products."

Keach is but one of many online marketers who are using site search results to squeeze better results out of their site and out of web marketing efforts. These data have always resided in the back end of site search; but now analytics make it easier to mine and use them. "The site search box is a survey taken every second," says Steve Kusmer, senior vice president and general manager of WebSideStory’s Search and Content Solution, a new offering founded on the analytics provider’s recent acquisition of site search provider Atomz. "It is, literally, finding out in visitors’ own words what they are looking for."

That’s a potential gold mine of information that marketers are using in an increasing number of ways. At Palm.com, an Atomz, now WebSideStory, client, data from site search led to a new merchandising initiative for a number of digital game products. Palm noticed a high volume of site search activity on certain games. It created a highlighted area that featured a group of the top-searched games and then displayed that group when any one of the games was typed into the search box. That feature increased conversions for the featured games by some 60%, according to Kusmer.

Driving a new strategy

At Backcountry.com, another user of the former Atomz product, data from site search has been a driver in a new strategy under which the online retailer of outdoor sporting gear is not just adjusting merchandising or marketing, but launching entirely new sites. Backcountry has built its brand by keeping its focus on high-end, high-performance equipment for the serious outdoor athlete. But for years, its record of failed searches—returning zero results because the site doesn’t carry a product—has shown that some people come to Backcountry to search for outdoor gear and apparel that falls outside that narrow focus.

The failed searches show what Backcountry could carry and likely sell, but expanding to a more mainstream selection might dilute the brand premise for its core customer, while the bigger customers of the specialized products might not go looking for them at Backcountry.com. The retailer’s solution has been instead to create new sites based in part on information it gets from what visitors look for but don’t find on the original site. Within the past year and a half, it’s grown a total of five new sites, focused on specialized concepts such as Backcountryoutlet.com for end-of-model gear, or niche product areas such as Dogfunk.com, a dedicated site for snowboarders.

Backcountry also uses site search data to recruit new vendors. It shows to targeted vendors whose merchandise it doesn’t yet carry a printout of all the failed searches on Backountry.com for that vendor’s products. "In some cases, those lists are in the thousands," says John Bresee, president. "You can tell vendors that you think the site has value, but there is nothing as convincing as bringing in the data and saying, ‘Here is the number of times people were looking for you on our site, but they bought your competitor instead.’"

More utility

Site search will have still greater utility for Backcountry in a few weeks with the planned launch of expanded search functionality that has the power to span all six sites. The idea behind the concept, which is called fall-through search, is that when a search on Backcountry produces no results or only a few results, the results listings will include the display of Backcountry’s other catalogs, identified as the company’s sister sites, and a few results from each. Shoppers can either click on those links to get to a sister site, or stay on Backcountry’s site, if they choose.

"Search on your site is the area where the most innovation is going on," says Bresee. "And it’s not about which provider has the best out-of-the-box search, but which one has the best professional service team that is going to work with you, quarter after quarter, to improve it."

Site search provider Mercado Software has refined closed loop merchandising around site search in three core areas and automated much of it. Customer Onlineshoes.com provides an example of how Mercado’s solution incorporates a feature called dynamic relevancy. "Type in women’s tennis shoes, and maybe 300 SKUs could come up. Equally weighted, with the same relevancy, that kind of information is not really useful to the average customer," says Bryan Surles, director of systems engineering at Mercado. Mercado’s tool breaks the tie by ranking the order of search results based on analytic data, such as conversions, gathered on the back end, meaning that the most popular shoes bubble up to the top of the listing. "That ties directly into increasing your conversion rate because it makes it more likely that someone is going to find those shoes, like them and buy them" says Surles.

Surles adds that what constitutes the most popular shoes changes every time the index is rebuilt, based on new data about conversions. Those change can occur as often as the retailer wants. But it doesn’t make sense to make those adjustments to search results in real time because that wouldn’t allow the retailer to see trends. "So every night, we take the web analytics report and repopulate search results based on what has been the most popular shoes in the trailing 30 to 90 days, tie that back to our relevancy ranking algorithms and optimize," he says.

More than algorithms

Conversions are just one element the tool can tie into site search results. "Search and merchandising need to be based on more than the fantastic algorithms we and our competitors have in our products," says Corey Leibow, CEO of Mercado. "We absolutely need to get to the next level by making use of real life factors such as buyer behavior and business drivers that we can now access in real time, over a period of time, including things such as profit margin and inventory levels, and put them into the business rules in an automated fashion."

Along those lines, another element of Mercado’s search solution, dynamic refinement, narrows a site’s left-side navigation tree according to analytic data and business rules in the same automated way. A third leg of the product can optimize the creation of site search landing pages based on the results of multivariate testing. Coupled with analytics results, the feature lets brand managers track test results over time, determine the best combination of page elements and with the click of a button, turn that combination into an actual landing page for specific search terms.

Endeca Technologies Inc. has developed an engine that can pull in data from its reporting tools developed for retail, which show how visitors interact with search, navigation, and merchandising on the site, as well as clickstream analysis. Though it has a partnership with Coremetrics Inc., is also integrates with other analytics providers. The engine can also pull data from other systems such as financial data; for example, which products have the highest profit margin. The combination is the basis of dynamic merchandising rules that can drive the activities of search, guided navigation and merchandising functions on the site in an integrated way that’s come to be called "searchandising."

Much bigger

Rich Stendardo, Retail Solutions Manager at Endeca, offers an example of how retailers can employ that functionality to meet business goals: If Home Depot cuts a new deal with supplier Delta Faucets, for instance, it could easily build a merchandising rule that would cover every search page on which Delta faucets appear, so as to feature the top-selling Delta faucets. Stendardo points out that Endeca’s engine can pull information that fits into templates, which puts the ability to change search and merchandising rules into the hands of business managers through Endeca’s Web Studio. A second Endeca studio provides added tools for IT staff and developers.

Search was once a discrete function on a web site with one role: to help shoppers get to the products they were looking for. And while that’s still uppermost in how consumers may understand search, behind the scenes, online retailers understand it’s something much bigger. Recast from passive guidepost to active merchandising and marketing tool and paired with the new power of web analytics, it’s a key part of "a whole-system view of visitor behavior, from the initial acquisition by a site, all the way through site search to conversion," Kusmer says.

mary@verticalwebmedia.com