American Bridal enjoys sales boost resulting from learning-based search system

From June 2007

By Lauri Giesen

American Bridal has been pleased with most of the features and functionality of the Yahoo! Internet sales platform for the 10 years it has been using them. The only real limitation was the search function: it was basic and generic.

When customers typed in a product name or description, they received dozens of listings for every product that contained that keyword or phrase. As a result, customers had to spend a lot of time sifting through the search results, many of which often were not relevant to their query.

The San Francisco-based online retailer knew it had to do something about the search feature, but it didn’t want to leave Yahoo! “Yahoo! was a great platform for companies that are just starting out or growing their business, but we knew we had to do something more with search,” says American Bridal CEO Shirley Tan.

While investigating alternatives, American Bridal found that a number of options did not work on the Yahoo! platform. Then it discovered Cupertino, Calif.-based SLI Systems. SLI uses “learning-based” search functionality, meaning that the system looks at what American Bridal customers have previously clicked on when performing similar search requests, and puts the most popular results on top.

Since implementing the system in August, sales revenue tied to customers using the search function has increased 400 percent. Overall sales for the site were up about 40 percent in 2006, and Tan attributes much of the gain to the improvement in search functionality.

American Bridal has found that the longer it uses SLI’s search functionality, the stronger the results become. “You need data to build on,” Tan says. With this system, the longer you use it, the better [the] data you can provide.”

Helping customers sift through voluminous options is important to American Bridal, which sells more than 2,000 different products associated with weddings — from bride’s and groom’s gifts to bubbles, candles and decorations. “We say we sell everything but the dress, and that is really true,” says Tan.

How did American Bridal know that its old search technology wasn’t generating the best results? Because its customer service agents were using it when shoppers would call in to inquire about products.

“We thought if the old search wasn’t very helpful to us, it could not be very helpful to our customers,” Tan says. Now, the results they are providing are much closer to what the customer wants, and “our reps can say to our customer, ‘This item is the most popular,’ and know that they are telling them the truth.”

Robust, reliable search functionality is extremely important to American Bridal. Since selling off its retail component in 2004, it receives all its business from online, phone and warehouse sales.

Bonus sales
And while the key objective of an internal search function is to help customers find the products they’re actively seeking, a good search engine can also show them products they are likely to want — but didn’t inquire about.

“If you give customers a good experience, you can help direct them to additional items that they were not originally looking for,” says Shaun Ryan, CEO of SLI Systems.

Using the SLI search, customers can sort results by popularity or highest/lowest price. When a basic term is given to the search, the system often requests greater specificity, allowing the customer to fine-tune her request before seeing the end results. Using the term “coaster wedding favor,” customers can specify if they want a photo coaster, a glass coaster or some other type.

Another feature: Customers interested in purchasing items featured in popular bridal magazines can request to see those products by listing the magazine title and an advertisement or editorial description.

But it is getting the most popular items in front of customers first that has really helped American Bridal’s sales. For example, a search for wedding returns 153 items. SLI’s research shows that the item listed first due to popularity had been No. 91 before American Bridal implemented the learning-based system.

“There is no way customers are going through 153 search results,” Ryan says. “Most customers will look at just the first couple of pages and, if they don’t find what they are looking for, they will assume the store doesn’t have it.”

The system is designed to identify common mistakes that customers make when describing products and “matches them up with the correct words used in the product description,” Ryan says.

SLI hosts the search function on its own server, but the results are presented to customers as if they were still on the American Bridal website. American Bridal pays SLI a monthly user fee based on the number of search queries.

“Charging by the number of queries allows us to set our fee based on a proportion to the value the retailer receives,” Ryan says. “The more people that use the search function, the more we charge, but most likely, the greater [the] benefit the retailer receives.”

Using an outside company to perform search function is fairly routine among small and mid-size e-tailers, and a growing number of large retailers are also beginning to outsource the search function. “Even the large retailers that have large IT staffs don’t have the experience and expertise specific to search,” Ryan says.

Next