• by Ray Schultz , Columnist, Yesterday

Email senders trying to discern who is likely to buy may have a new tool: an artificial intelligence-based model that “predicts digital customer behavior and delivers personalized marketing insights across complex, multi-touchpoint journeys — outperforming traditional methods in both precision and ROI,” according to researchers at the University of Maryland’s Robert H. Smith School of Business.

Their findings will appear in the Journal of Marketing Researcher under the title: “AI for Customer Journeys: A Transformer Approach” 

The new approach applies transformer-based models developed for language processing to analyze multichannel sequences of customer interactions, the researchers claim. 

“When analyzing a sequence of customer interactions, it is important for firms to understand how these interactions align with key objectives, such as generating qualified customer leads, driving conversion events, or reducing churn,” an abstract of the paper states.

The authors propose “a transformer-based framework that models customer interactions in a sequence similar to how a sentence is modeled as a sequence of words by Large Language Models.

As with many academic projects, it is not yet clear what the real-world results will be.”  

“Transformers give us the ability to see the journey as a whole, not just as a series of isolated interactions, says P.K. Kannan, dean’s chair in marketing science at the university and co-author Zipei Lu, a PhD candidate “That’s a major leap in marketing analytics.” 

Unlike traditional journey methods and LSTMs, Hidden Markov and Poisson Point Process models, Kannan and Lu say their approach “captures both the timing and nature of each touchpoint, making it ideal for today’s fragmented, multi-touch marketing environments.”

A central contribution of the paper is the integration of customer-level heterogeneity within the transformer architecture. This allows the model to deliver individualized insights into how different customers respond to marketing actions over time.

“We designed the model to capture the complexity and individuality of digital customer journeys—something traditional models often overlook,” says Lu.

Kannan adds, “Incorporating customer heterogeneity allows us to move beyond one-size-fits-all journey maps. We’re now able to understand how different customers respond over time—and act on it.”

The resulting model “doesn’t just tell us who’s likely to convert. It tells us why, and more importantly, when to act,” Lu adds. 

The authors studied journey data from a large hospitality firm, covering over 92,000 users and over 500,000 touchpoints.