Spiraledge taps natural language technology from RichRelevance Inc. to improve product recommendations and generate "good friction" for online shoppers.

Increased competition for web traffic and the rising costs of customer acquisition continue to make ecommerce more challenging. To mitigate those problems, Spiraledge Inc., the parent company of retailers that sell swimwear and yoga gear and apparel, is using artificial intelligence (AI) technology to make the most out of every consumer visit.

“Retailers are forced to maximize conversion rates and average order volume (AOV) for the traffic they have,” says Alexander Sienkiewicz, chief marketing officer at Spiraledge, which operates sites SwimOutlet.com and YogaOutlet.com. It ranks No. 256 in the Internet Retailer 2018 Top 1000.

Alexander Sienkiewicz, chief marketing officer at Spiraledge.

To help it make the best of each consumer visit to its website, the retailer about five years ago started using technology from personalization vendor RichRelevance Inc. to help guide consumers through its ever-changing catalog of more than 100,000 SKUs.

“The more SKUs a retailer has, the more important personalization technology is,” Sienkiewicz says. Without personalized recommendations, consumers easily can get lost and confused as they wade through thousands or tens of thousands of potential product options, he says.


In addition to helping consumers find what they want, the technology can be used to surface related merchandise relevant to each consumer—potentially increasing the number of items in each shipment and AOV, Sienkiewicz says. That’s important because boosting the number of items in each purchase is more profitable than selling the same amount of goods in separate orders, he says. That’s in part because the retailer only has to pay to acquire one customer.

For each of those tasks, the personalization technology has worked well but with certain limitations, Sienkiewicz says. Conventional personalization technology builds recommendation data based on the clicking, browsing and buying activity of consumers—and that can limit its effectiveness for retailers like Spiraledge, which continuously adds large numbers of new items to its catalog, he says. Because new items come with no behavioral data, they might not immediately show up in the product recommendations of consumers, leading to lost sales opportunities, Sienkiewicz says.

To address that problem and make its personalization work better, Spiraledge recently became an early user of RichRelevance’s Hyper-Personalization technology, which uses natural language processing (NLP), a technology that allows computers to understand and analyze human languages.

The new RichRelevance technology is designed to generate product recommendation data faster. Like other technology, Hyper-Personalization looks at behavioral information, but also uses NLP to analyze text information, including such things as product descriptions and manufacturer-provided information or product reviews that appear its website or others, Sienkiewicz says. That NLP analysis allows the technology to learn faster than older methods which relied heavily on consumer interactions, getting each new item in front of appropriate consumers more rapidly than other approaches, says Raj Badarinath, vice president of marketing and ecosystems for RichRelevance.


For example, a shopper who searches for apparel with unicorns might be offered items with similar qualities such as “soft” or “fuzzy,” even if the consumer didn’t enter those search terms. The Hyper-Personalization product can find and use attributes described in any of the information about text information analyzed by the NLP. The technology compares those attributes to those of items the consumer is already looking at—in real time, Badarinath says.

“This is a clear departure from the use of behavioral data,” Badarinath says.

‘Good friction’

Beyond generating quicker recommendations, a goal of the personalization technology is to make the shopping and buying process easier, but in some cases, also a bit slower. The RichRelevance software seeks to facilitate a kind of “good friction” by showing consumers more relevant product recommendations about more things they want or need, Badarinath says. Slowing down the process just a little, he says, can prompt consumers to order more items, potentially resulting in larger orders.


That is important for smaller retailers like Spiraledge, Sienkiewicz says, because it’s much easier to make money on orders that contain at least two items. “We want to maximize our volume per order,” he says. Personalized product recommendations helps the retailer do that, he says.

On average, Spiraledge sells more than two items per order, Sienkiewicz says, without providing an exact number.

The implementation of the new RichRelevance technology is still in the early stages, but initial results have been promising, Sienkiewicz says. “Directionally, we’re seeing things going in the right way,” he says. For example, Spiraledge saw an 11% increase in click-through rates of recommended products. It’s too early to assess what impact the NLP technology will have on sales, he says.

Including Spiraledge, RichRelevance provides personalization services to 37 retailers in the Internet Retailer Top 1000, according to Top500Guide.com.