In the next 17 months, it expects 10% of its B2B customers will be transacting on the web, an executive says.
Although site search is a major component in the performance of any retailer’s web site, it nevertheless remains one of the most underutilized merchandising tools by retailers.
The chief shortcoming of site search is the inability of the application to match partial search strings to a desired product or to recognize new words in the shopper’s vocabulary for an item in stock. The latter is a problem that is far greater for apparel retailers, which typically offer several shades of the same color.
Consequently, shoppers receive results with low relevancy to their query or no results at all, a.k.a. 0% search. Such poor site search experiences translate into lost opportunities for retailers to convert a shopper into a buyer, as well as engage in cross-selling and up selling.
“Site search is about more than site navigation, it’s about optimizing conversion rates, as well as cross-selling and up selling,” explains Corey Leibow, president and CEO of Mercado Software Inc.
Uprooting a home-grown system
Recognizing that its site search and navigation applications did not work together in a relevant manner, thus suppressing conversion rates and the number of shoppers that used its site search feature, apparel retailer Guess? Inc. enlisted Mercado in 2005 to help revamp its web site with the goal of re-launching in time for the 2005 holiday shopping season.
“We had a homegrown site search application and the relevancy of the results was weak,” says Guess? director of e-commerce Michael Africa. “Customers that typed in partial search strings got hits about 30% of the time. Research told us if we could improve the search function it would lead to higher customer satisfaction, conversion rates and sales.”
Mercado attacked the problem by first gaining a thorough understanding of Guess?’s business objective, which was to improve conversion by delivering search results that enable customers to navigate more easily to pages containing items with the desired color, size and price point. “We saw a client that needed to walk before it ran, which is why we took a phased approach to implementation,” says Leibow.
First, Mercado addressed the shortcomings in Guess?’s site search application by loading synonyms and related terms into the search thesaurus for each item in the Guess? Catalog. That effort made search more intuitive by broadening the array of colors and style variations. Mercado also replaced Guess?’s internal search engine with one that also included a business rules engine – featuring an algorithm that returns like items for words or partial strings that do not return a hit.
This feature proved a major upgrade from a merchandising perspective, because it allowed Guess? to push product to shoppers under any circumstance, something which its previous site search application was incapable of. “The one thing search should never do is return zero results, even if there is no match,” stresses Guess? chief information officer Michael Relich. “At the very least, we needed to show best sellers in the event of a non-match to keep customers engaged.”
Having beefed up Guess?’s site search dictionary, Mercado began addressing larger merchandising issues by linking search to site navigation. Many shoppers type in a single search word and navigate the site based on those results without typing in another search word. But if the results aren’t what the user was seeking, that approach can lead to a dead end if the site’s navigation bar remains static. A static navigation bar offers no alternatives to shoppers based on what they were looking for and makes it difficult for shoppers to drill down to locate an item by color, price, size or other criteria as they move through the site.
Mercado addressed the problem by developing a navigation bar that changes in real time based on the shopper’s search results and subsequent movement through the Guess? site. The navigation bar also changes accordingly with each new search word entered. The new application makes it easier for shoppers to locate pages with the desired items and zero in on the item with the exact features desired by the customer.
“Each shopper likes to search differently, and allowing search and navigation to work together in real time creates greater relevancy of results and higher customer satisfaction,” says Leibow. “Enhancing site search without integrating it to site navigation does the retailer’s business, and the shopper, an injustice.”
More searching, more sales
The payoff for Guess? has been a more than three-fold rise in the use of site search and a more than five-fold increase in conversion rates among customers who use site search. Overall, conversion rates for the site, which was re-launched in early November 2005, are up 50%. Customer satisfaction during the holiday shopping season jumped to a rating of 79 out of 100, on average, up from 77 the previous holiday shopping season, based on industry standard measurements. By contrast, Relich notes, “A lot of retailers saw a one- to two-point drop year-over-year during that period.”
The results have been so stellar Guess? has rolled out the new site search and navigation features to sister sites Marciano.com and Guessfactory.com. Both sites have experienced similar improvements in conversion rates, use of site search and customer satisfaction, according to Relich.
Guess? now receives daily reports from Mercado detailing site search results and subsequent navigation patterns, which the retailer uses to create highly targeted e-mails promoting items that show the highest click-through rates. The promotional pieces are sent to customers who have shown interest in similar products without making a purchase. The e-mails include links to dynamic landing pages capable of delivering higher conversion rates.
Having improved the logic behind the business rules powering its site search and navigation tools, Guess? has begun applying the power of Mercado’s software platform to creating time-sensitive banner ads that run for a specific period and are then replaced by another time-sensitive ad.