Ulta Beauty embraces low-code to deliver better CX

About six years ago, Ulta Beauty formed a dedicated innovation team to identify technologies that resonate to improve the customer experience. As the VP of digital innovation, Michelle Pacynski heads up this team and oversees around 40 innovators whose job is to leverage new and emerging tech to come up with fresh and creative value propositions for the business. Over the years, they’ve created a virtual make-up try-on tool using augmented reality, played around with intelligent mirrors, and used AI to build their personalization engine, which intelligently mines customer data to give product recommendations.

They’ve also been using low-code and gen AI to quickly conceive, build, test, and deploy new customer-facing apps and experiences. In particular, Ulta utilizes an enterprise low-code AI platform from Iterate.ai, called Interplay. Offering hundreds of different products and services from a variety of startups that can be easily dragged and dropped into the Interplay interface, people can easily test how different components work, she says. The goal is to experiment quickly and identify solutions that appeal to customers. In a fiercely competitive industry, where CX is critical to differentiation, this approach has enabled them to build and test new innovations about 10 times faster than traditional development.

[ Learn 10 tips for a successful CX strategy and what it takes to provide best-in-class customer experience | Sign up for CIO newsletters! ]

Flexibility, scalability, searchability

Low-code also makes experimenting less risky. “Not only does this particular low-code solution make rapid experimentation possible, it also offers orchestration capabilities so we can plug different services in and out very quickly,” says Pacynski. “And if it doesn’t work, we have the flexibility to take out a component and put it in something different without any hassle.” Plus, this strategy has enabled them to easily move from experimentation to production, speeding up the process of transforming their initial ideas into fully deployed applications. Using Interplay, it’s seamless for Ulta to string together all the plumbing behind customer experiences using the best solutions for each unique use case.

As an example, Pacynski says they used Interplay to build their customer service chatbot about four years ago. Initially, it was designed to listen to conversations on social channels, like Facebook and X, but they quickly realized customers didn’t want to go to a social channel to resolve issues. With Interplay, it was relatively simple to take what they already created and migrate it onto the website. And now, with the rise of gen AI, they can evolve further because the chatbot has so many more use cases than what they originally intended.

As the biggest beauty retailer in the US, it’s critical for Ulta to use technologies that can quickly scale. “This low-code solution is also the foundation of our personalization platform,” she says, “so it serves as the base for providing our customers with product recommendations. It must be able to serve up millions of recommendations every day.”

Product search is also incredibly important, but equally complicated to get right. If a customer searches for “light red lipstick,” for example, there could be hundreds that are essentially light red but not labelled as such, and so their customers won’t find them. Ulta used low-code to build out new data models that make is possible to better map different product search words and plug in custom components, as well as off-the-shelf components, to simplify what customers are looking for.

While one could assume their capabilities have dramatically improved with the rise of platforms like ChatGPT, Pacynski is quick to point out these tools aren’t mature enough to do what they need them to do. And where they can deliver, she says, they need to be used strategically. For example, if a customer is looking for the best moisturizer for someone with oily skin that doesn’t contain certain ingredients, it’s important the algorithm only pulls results from the Ulta product catalogue and doesn’t suggest products they don’t carry in their stores. This kind of thing needs to be part of the product scoping process, she says.

Securing business buy-in

Developing applications in this way may have been part of the innovation team’s DNA for several years, but taking this approach to the enterprise hasn’t been easy. “It’s a journey we’re still on,” Pacynski says. “We’ve had many conversations about low-code tools broadly in the enterprise, but there’s a lot more planning, thought, training, and education needed to bring in new ones, especially because these teams already have existing solutions and methodologies.”

One way to convince them about the value of this approach is to showcase results. “The omnichannel strategy at Ulta has been strong for many years,” she adds. “We’ve had buy online, pick up in-store for a long time but we didn’t offer buy online, pickup curbside. During the pandemic, we used Interplay to build an interface to make curbside pickup possible across our 1,400 stores in a couple of weeks.”

When you’re really trying to create a competitive advantage, explore new territories, and do something that differentiates you from the competition, low-code allows you to iterate quickly, and play around without the risks, she says. But this doesn’t mean you can just test forever. “With any project, you have to set out clear development timeframes so experiments don’t run too long without yielding any tangible learnings or results,” she says.

More on improving the customer experience: