Despite a burdensome number of search rules, a sophisticated athletic-wear brand’s Endeca-powered on-site search engine couldn’t provide its customers with the distinguished experience they expected and received in-store.
Translate the brand’s welcoming in-store atmosphere to its website using signals to personalize guests’ search and browse experience while reducing the number of search rules for its team to upkeep.
With Fusion deployed, click-to-cart increased by 28% on Black Friday compared to the previous year. Simultaneously, query response times were reduced by 50%, and search rules were reduced by 97% from the Endeca engine.
Customer experience is at the heart of one of the world’s most popular activewear brands. In its stores, the brand provides customers (who it calls “guests”) with premium service via store “educators” (sales associates). Guests arriving at a store are given as much or little service as they desire. Each store is custom designed by its educators, localizing the look and feel to the geographic region for a more personalized and intimate shopping experience. Classes are offered to support and encourage customer interests. Guests are made to feel unique, honored, and valued.
The brand had a vision of translating its first-class in-store experience to its website. Inspired by other retailers with exceptional online shopping experiences (Target, Walmart, Amazon), it identified several ecommerce features required to achieve this goal: security, intuitive design, mobile optimization, payment options (credit card, Apple Pay, Google Wallet), order tracking, product reviews, and, finally the shopping experience itself — search, browse, and navigation.
At the time, the brand’s search engine was built on Endeca, which was employed across its website. The homepage, search, browse, and product detail pages all relied on Endeca. Using Endeca, each browse page had to be individually sequenced by a search team member, meaning each product type required someone to specify which particular product should be first on the list, which should be second, third, fourth, and so on.
When the out-of-the-box search experience still wasn’t satisfactory, rules had to be applied to make the Endeca engine perform the way the brand preferred. Manually-created rules were used to facilitate synonyms, create redirects, etc. For the roughly 1,500 products offered, more than 3,000 search rules were used to curate a satisfactory experience. That’s at least two rules for each and every product.
The amount of human intervention and manual manipulation required with Endeca was simply not scalable. Nor was it possible for the brand to provide the desired, distinctive experience to its online guests so long as Endeca was under the hood.
Of course, the brand wasn’t making things easy on the search engine either. Rather than being descriptive, product names were catchy. Names like “Hip Hippie Shorts,” “Set To Jet Pants,” and “Set To Sweat Pants” didn’t give the search engine much insight into actual product features like fabric, fit, and intended activity.
Product descriptions, just as colorful as the names, provided little help, leaving the search engine short on data to index and upkeep demands of the search engine high.
The activewear brand knew it needed a new search tool that was enhanced by automation, machine learning, and a flexible UI to support a seamless UX. The team wanted to use its own homegrown ML algorithms written by a team of data scientists to customize the search and browse experience in automated, powerful, and sustainable ways.
The search team could build a custom-designed engine from scratch, but capabilities that came off the shelf with some of the pre-built search platforms would take years to replicate on their own.
Which search tool was right for the needs of the sophisticated, customer-centered brand? The market for on-site search engines has a dizzying number of options so choosing the right one took some research. Sifting through the features and functions of all the possibilities, Lucidworks Fusion rose to the top.
The uniqueness of the brand’s customers and products matched nicely with Fusion’s customizable infrastructure. In addition to its powerful, out-of-the-box machine learning capabilities, algorithms created just for the brand’s shoppers could be plugged into the system. A search engine with a black box of machine learning wasn’t what this manicured brand wanted.
Fusion’s Predictive Merchandiser add-on gave the brand’s merchandisers the power to perfect the shopping experience without needing to rely on IT. With an easy-to-use visual interface, merchandisers can intuitively curate product placement and search results based on their expertise so online shoppers get the in-store quality they’re used to.
Beyond search platform capabilities, the brand also wanted a partnership that supported its move to delight guests online; someone it could rely on for best practices and implementation help. Lucidworks’ extensive knowledge of crafting search and browse experiences across multiple channels for many of the world’s top retailers was just what the brand needed to succeed.
The new search and browse experience had to be all buttoned up before the big Black Friday holiday rush. After purchasing it in February, the brand set a goal of getting Fusion live on its website in the summer.
The team began to pull together the data feeds Fusion would rely on for indexing. On Endeca, the catalog feed ran only a full, baseline index. This was a big disadvantage since it meant real-time changes based on inventory availability, new products, and so on were not possible. For better website performance and user experience, Fusion enabled incremental updates to the catalog data.
The inventory feed coming into Fusion would include not only the ecommerce products Endeca had served up, but also products available only in retail stores to support a true omnichannel shopping experience. Additionally, inventory quantities, a feature not used with the Endeca search engine, would also be included in Fusion’s data.
Some scattered signal processing had been done previously around product revenue, but Fusion allowed for a more global extraction and view of signals. This data would then be married with analytics in Adobe Experience Manager.
Before the brand deployed its upgraded on-site search and browse to the masses, it had a number of objectives to meet:
In August, the new Fusion-powered search was ready for primetime. To allow for testing, 5% of search and typeahead traffic was initially sent through Fusion. Then, as Fusion raced through all the required gates, traffic was steadily increased to 10%, 20%, 50%, and finally 100%. Browse, menus, and recommendations were also turned over to Fusion’s care.
The more traffic thrown at the system, the better performance got. More traffic enabled better analytics and allowed the brand to more finely tune and improve search and browse. When traffic reached four-times the website’s peak traffic of two years earlier, response times with Fusion far outperformed the previous numbers. Fusion reduced query response times by 50%. Behind the scenes, the continuous updates enabled by Fusion reduced batches from fourteen per week down to one.
In addition to a rewarding experience for shoppers, improvements to the search system benefitted the brand’s employees. Those 3,000 rules the search team had to tame with Endeca were reduced to a more manageable 100 with Fusion. Staff could turn to more engaging and impactful work instead of monotonously adding one thesaurus entry after another.
Signals-based sequencing in Fusion, with a few business rules layered on top, replaced the manual product-by-product sequencing Endeca required.
Another big benefit of Fusion was the ability to experiment using A/B testing. This allowed standard ideas and practices to be challenged and either confirmed or refuted theories using actual data. Maybe women’s clothing should not always appear above men’s (an assumption based on the brand’s early beginnings in women’s apparel and the likelihood of the shopper, therefore, to be a woman). Maybe discounted products should be shown above full-priced items to benefit cost-sensitive guests. None of this experimentation had been possible with Endeca.
As the brand got more sophisticated with its Fusion implementation, it began to provide guests with personalized online experiences fit to their interests. Using Fusion signals, its website is automatically curated to the individual, echoing the personalized interactions an educator would provide in-store: suggesting products they’re likely to be interested in, ordering suggestions based on activity, color, or gender preferences.
Fusion also provides a window where the brand can observe the language its customers use to hunt for products. By analyzing search term data, it can more easily speak the language of its guests when describing products to meet shoppers where they are. By using customers’ own vocabulary, the brand makes guests feel at home on its website.
Switching to Fusion meant the brand no longer had to provide the best average experience to all of its guests and instead offers easy, personalized pathways for each person to reach their specific goals in a manner that makes them feel known and understood.
When Black Friday came knocking and searches on the brand’s website ratcheted up to 2,000 per second, Fusion kept the responses rapid, leading to 27,500 orders per hour. Click-to-cart rates increased 28% over the previous year, proving the Fusion-powered search system understood the brand’s guests and knowledgeably guided them to the right products, just like an in-store educator.
