Lucky Brand accents site sales with product recommendations

Product recommendations are a proven way for web retailers to drive cross-sell and up-sell revenue by promoting additional merchandise to site visitors. At Lucky Brand, an accessible premium denim lifestyle brand with a complete offering of apparel and accessories, marketers are making the most of the company’s IBM® Product Recommendations implementation with ongoing optimization and a strategic initiative to take full advantage of the technology’s capabilities.

A Liz Claiborne Inc. subsidiary, Lucky Brand uses Intelligent Offer as a key part of its digital marketing program to increase personalization and engagement with its fashion-savvy, brand-conscious customers. The company has learned from experience that it pays to analyze and optimize recommendations to match the shopping behavior of its customers.

“We have a focus on testing and tweaking business rules in IBM Product Recommendations to identify which business rules work best for our online store and where we can get the greatest increase,” said Sanam Jivani, Lucky Brand ecommerce manager. “Intelligent Offer is definitely driving incremental revenue for us. It’s a very intuitive and robust tool and cross-sales have improved as a result.”

Intelligent Offer drives up to 16 percent of site sales
With IBM Product Recommendations, the retailer, founded in 1990, has seen recommendations-influenced sales rise as high as nearly 16 percent of total site sales, a strong number that underscores the returns that merchants can realize from the solution and a focus on continuous optimization to deliver personalized interactions.

For example, as part of a website redesign, Lucky Brand decided to move product recommendations to the right column, away from their previous placement at the bottom of pages, “below the fold.” The move
was endorsed as a best practice by IBM Coremetrics consultants engaged by Lucky Brand to help support its digital marketing efforts.

“We wanted a clean display of four recommendations on the right side of every product page allowing them to be more visible. It’s definitely a better placement—we’re seeing that in the positive results!” Jivani said. Shoppers agreed, as the percentage of site sales influenced by Intelligent Offer leapt to about 14 percent in just one week—more than double its previous average—and rose further in the weeks to follow.”

In another bid for optimization, Lucky Brand experimented with custom business rules to govern which products are recommended with which products. With its default “wisdom of the crowds” algorithm, Intelligent Offer will display items similar to the merchandise a customer is browsing—for instance, a customer browsing jeans will see similar jeans viewed by similar customers under the heading, “You May Also Like.”

With a custom approach to “outfitting” the customer, Lucky Brand tuned Intelligent Offer to display complementary items—a matching blouse for a shopper viewing jeans, for example. Marketers reasoned the “outfitting” approach would mimic the experience in one of Lucky Brand’s 217 stores in North America (including outlet and retail) or at select retailers selling its products, with sales personnel recommending items for a matching outfit.

“Initially, we provided recommendations by showcasing an outfit. For example, if they were on a jeans product page, we wanted to recommend a top, bag or shoes that would look good with those jeans,” Jivani said. “But it turned out that if the customer was in the denim section of our online store, they wanted to see more denim. It proved to us that the Intelligent Offer algorithm worked better than trying to force a way of shopping.”

With the experiment, Lucky Brand learned the most effective technique for recommendations and gained experience in how an enterprising approach can contribute to optimization.

Greater flexibility with dynamic recommendations
To fully leverage Intelligent Offer, Lucky Brand has moved up to deploying dynamic recommendations, which offer greater flexibility, accuracy and opportunities for optimization, including real-time A/B testing, than the flat file recommendations previously in use. Lucky Brand may use the new A/B testing capabilities to revisit offering “outfitting” recommendations, re-evaluate results and further optimize its recommendations strategy.

“Using dynamic recommendations gives us a lot more flexibility than we had with flat files,” Jivani said. “Now we can do A/B testing, 50/50 splits, test on zoning, content placement, header text and more. It gives us more capabilities to better understand our customer and what they react to while shopping on the site.”

Moving to Intelligent Offer dynamic recommendations also automates processes and eliminates issues Lucky Brand occasionally experienced with the flat file method. In the past, Intelligent Offer would export a large, static flat file of recommendations in a daily batch to a third-party vendor that would post them to the site. However, the process did not always go smoothly, as recommendations would occasionally be missing or not updated.

With the dynamic model, the Lucky Brand storefront makes a real-time call to Coremetrics to fetch recommendations on the fly, eliminating a point of potential failure by integrating directly with the website. “Since we’ve been with dynamic, I always get email notifications that the feeds have been successful,” Jivani said. “We’ve haven’t seen any breakage in the system that would require us to take time to troubleshoot.”

A continuous focus on optimization and personalization
With dynamic recommendations, Lucky Brand can refine A/B testing to assess recommendation strategies by gender, and tie recommendations to search terms. It’s also exploring introducing a greater degree of personalization into the solution, with recommendations based wholly on browsing activity or in part, along with “wisdom of the crowds” algorithms.

Another area on the drawing board is adding recommendations to email communications. For example, Lucky Brand can use Intelligent Offer to feature products in retargeting emails it sends to customers who abandoned items in a shopping cart.

Core to Lucky Brand’s recommendations and website optimization initiatives is its use of the IBM Digital Analytics platform, which the retailer used extensively in a site redesign. Lucky Brand also uses the solution to measure the effectiveness of its digital marketing campaigns, including how customer interaction through recommendations, email and paid search ads influence conversion.

“IBM Digital Analytics is a very robust solution that gives our online marketing and merchandising teams the data we need to drive online sales and satisfy our customers,” Jivani said.

For more information
To learn more about IBM Coremetrics, please contact your IBM marketing representative or IBM Business Partner or visit the following website: ibm.com/software/marketing-solutions