How Therabody Uses Connected-Product Behavior To Reveal What Consumers Actually Need

AI is reshaping the path to purchase in real time. Consumers now routinely ask large language models what products to buy, compressing discovery, recommendation, and evaluation into a single interaction. Simultaneously, the broader e-commerce market continues to move away from aggressive growth tactics and toward profit-focused operating models built around retention and efficiency. Against that backdrop, Therabody is refining how it interprets consumer behavior. The company now uses connected-product data to close the gap between what consumers claim they want and what their real-world routines actually reveal over time.

Navigating that environment requires a much deeper integration between marketing, software, and product development. Caitlin Berzok, SVP of Digital at Therabody, approaches consumer growth through that lens after nearly two decades working across digital product and experience design. Before joining Therabody, Berzok helped build JPMorgan Chase’s first gamified financial-health product, experience that shaped how she thinks about consumer behavior inside connected ecosystems. She originally joined Therabody to lead the company’s software and mobile organization, a background that now heavily influences how the brand manages growth across its e-commerce, app, and connected-device business.

"You can survey people all day long, but what people say versus what they do is very different. There’s value in both, but you need to put them together to really get the insight," says Berzok. That mindset now shapes how Therabody approaches visibility inside AI-driven commerce environments. Berzok says PR coverage, affiliate mentions, and earned media function as signals not only for consumers, but also for large language models that now shape product discovery and recommendations. Many shoppers begin their search process through tools like ChatGPT or Claude rather than traditional search engines, forcing brands to think about discoverability across both human and machine audiences simultaneously. "You've got to market to the humans and market to the robots," she adds. "We're in a different world. It is not necessarily a linear path anymore where the consumer sees one thing and then sees another."

AI's layer of insight

Inside Therabody, the company’s profit-focused growth model relies heavily on AI to turn fragmented consumer signals into usable operating insights. Berzok says workflows that once required manually weaving together spreadsheets full of reviews, ad metrics, surveys, and social chatter can now be synthesized much faster through large language models. Her team regularly runs large datasets through Claude to identify patterns product and marketing teams can actually act on, dramatically reducing the time previously spent sorting through raw customer feedback. "Being able to synthesize and stitch together all of that information to identify trends was historically a very manual and cumbersome process," Berzok notes. "It was a little bit of art, a little bit of science, and a little bit of hope to see if it all came together."

Connected-product behavior gives the company an even more detailed layer of insight underneath those broader datasets. Berzok says first-party behavioral data often reveals patterns consumers would never explicitly articulate in surveys or reviews alone, particularly as users interact with connected hardware and the Coach by Therabody recommendation engine. "We start to build a picture of, 'What are the most common activities that our users are doing?'" Berzok says, describing how the system combines wearable inputs, activity data, and recovery goals to surface emerging behavioral trends. In one case, Therabody identified a strong concentration of yoga-focused users, an insight that later shaped the recovery content and product-detail experiences shown to consumers searching for mobility support around specific poses and stretching routines.

The personalization tightrope

As Therabody gathers richer behavioral data, Berzok says the challenge moves toward using it responsibly. Consumers now expect fast, personalized recommendations and seamless digital experiences, especially as AI tools normalize tailored interactions across commerce platforms. At the same time, personalization can quickly become uncomfortable when brands overreach or surface information in ways consumers did not explicitly invite. Berzok says modern consumer expectations now sit in a delicate middle ground where relevance is valued, but excessive familiarity can easily undermine trust. "Personalization has become table stakes in a lot of ways. Consumers expect that personal interaction," she notes. "With big swings, you can go wrong very quickly. There's always a creepiness factor with how personalized something is. People want you to know them, but then they don't want you to know them."

To navigate that tension, Therabody focuses heavily on transparent, opt-in experiences that allow consumers to control how much information they share with the brand. Berzok pointed to tools like the company’s product quiz as examples of personalization that feel collaborative rather than invasive, helping consumers voluntarily guide the recommendations they receive across the platform. In her view, the goal is to use behavioral insight in ways that improve the customer experience without making the interaction feel overly engineered or intrusive. "At the end of the day, our products are going to make you feel better, sleep better, or look better, and the best way we can help you is through what we know about you."

Machine-speed leadership

For Berzok, the larger leadership challenge surrounding AI has less to do with adopting individual tools and more to do with continuously reexamining how teams operate. She says marketers can no longer afford to rely too heavily on assumptions built during earlier eras of digital growth, especially as consumer behavior and discovery patterns continue evolving at high speed. To pressure-test decisions, Berzok often uses frameworks like the "Five Whys" to unpack why something worked historically and whether the underlying logic still holds up in the current environment. "Expect change and expect that you will need to adapt. Even if you've been doing this for almost 20 years like I have, you need to challenge the assumptions that you had about what works," she advises. "The best practices of six months ago, five years ago, or 10 years ago simply do not work in today's day and age."

As the AI ecosystem continues maturing, Berzok believes the brands navigating the transition most effectively will be the ones pairing experimentation with strong operational guardrails. She describes her management style as a player-coach model, staying closely connected to the tools and workflows her teams use while also creating systems that allow employees to experiment safely without putting customer trust or business integrity at risk. That balance between automation and human oversight sits at the center of Therabody’s broader operating philosophy, where AI functions less as a replacement for strategy and more as an accelerant layered on top of disciplined decision-making and strong teams. "I like to think of it as gasoline you're pouring on the team that allows the brilliant people we have working for us to do their jobs easier and take some of the busy work off their plate," Berzok concludes.