Coconut oil-based beauty brand Kopari has gotten more serious about customer data and, subsequently, messaging its products.

Though the 3-year-old San Diego-based company has utilized A/B testing to cull consumer preferences since its 2015 inception, it has taken an AI-powered approach in recent months with the help of personalization technology company Dynamic Yield.

Following the January 2017 influx of capital from L Catterton and celebrities, like Ashton Kutcher, Mila Kunis and Karlie Kloss, Kopari began working with Dynamic Yield on personalizing its homepage and product page layouts based on how the customer came to the website and on what kind of device. The goal, said Megan Whitman, Kopari chief digital officer, was to give every consumer a custom experience on the site.

Whitman found that loyal customers, who are repeat purchasers and were coming to the site to scour new products and check on their Kopari Clique rewards points, didn’t need an intro to the brand, like first-time customers. Thus, Kopari loyalists now see promotions for its new lip scrub, while newer customers are being messaged marketing around its aluminum-free deodorant, which launched in January 2018. From Whitman’s perspective, the new customer needed to be informed about the latest product announcement information, as well as the brand’s story.

By tailoring messaging to both of these customer segments, Kopari has increased average revenue per user by 60 percent. Customer conversion is also up, having increased 30 percent year-to-date for the January to September time period. According to the company, repeat customers make up 50 percent of its database, and new customers are also 50 percent.

Whitman also has found through this AI testing experiment that as the company has expanded from just four multitasking products (those that can be used on the face, hair and body) to other categories like personal care and body care — Kopari now has 25 items in its assortment — customer purchase behavior has taken a category-centric bent.

“People who have bought body care before are more interested in buying more from that line than, say, potentially buying skin care or personal care,” said Whitman. “Through this personalized testing, we’ve been able to guide customers’ shopping experiences and drop them into exactly what they are looking for and want. It makes it easier for them to actually purchase.” Fittingly, repeat customers rarely visit the brand’s homepage, said Whitman.

Dynamic Yield did similar product-detail page personalizations for Sephora in Asia, and its recommendation engine powered a click-through rate of more than 4 percent, returning direct revenue in excess of $6.50 for the retailer.

For Kopari, these learnings have helped grow its subscription program, which also launched in January and now makes up 20 percent of overall revenue. While, at first, the same subscription service marketing was applied to all products, Kopari has now tweaked this.

Based on consumer data, customers were subscribing to its deodorant at a higher rate than any other items, so now Kopari has made the default option on the deodorant subscriber-based. “Learning this, we adjusted the on-site experience to support that,” said Whitman. However, if customers shop on the Coconut Melt page, for instance, it doesn’t automatically drop them into the subscription model.

It works the other way, too. When Kopari launched its Shower Oil this month, it quickly landed in the top-three most-purchased products, making it ripe for subscription. Kopari has now employed this service for that single product.

These in-house AI learnings are being utilized by Kopari’s retail partners, as well, like Sephora, Ulta and Anthropologie. When Kopari noticed that there was higher engagement around its aluminum-free deodorant education, the brand created custom PDFs for Sephora to include in its own Kopari product descriptions.

“Even though we can’t directly test on Sephora’s site, we are able to reflect our learnings to them,” she said. The new Sephora product descriptions, complete with the Kopari-supplied PDFs, roll out on this week.

Fifty percent of the consumer tests that Kopari runs do not align with the brand’s gut reaction. Still, the company is continuing to run up to 10 AI and A/B tests every other week. “We always want to respond quickly to what’s working. That’s what’s going to keep us ahead,” said Whitman.