When companies try to predict product sales, they often look at one product at a time. But according to new research co-authored by Huimin Zhao1, that approach can miss a much bigger story about how customers actually shop.
Zhao, a professor of Information Technology Management in the Lubar College of Business, studies how relationships among products influence sales. His recent research shows that forecasting improves when businesses move beyond the traditional idea that products are either “complements” or “substitutes.”
“Most models assume products only matter to each other in very limited ways,” Zhao said. “But in real life, customers make choices based on many signals at once.”
Classic examples of complements include phones and phone cases, while substitutes might be two similar smartphone brands. Zhao’s research shows there are many other connections that matter. Products can be linked because they are displayed near each other, share a brand reputation, or follow similar buying patterns over time. Some relationships are indirect or one-sided, and others change as technology and consumer habits evolve.
“Product relationships are not fixed,” Zhao said. “They can grow stronger, weaker, or even flip from complements to substitutes as markets change.”
To capture this complexity, Zhao and his co-authors developed advanced artificial intelligence models that analyze sales data across many products at once. These models learn how products influence each other over time and use that insight to make more accurate forecasts.
The results were striking. In tests using real retail data, the new approach consistently outperformed existing forecasting methods. It also uncovered surprising connections. In one case, sales of a respiratory drug and a cardiovascular drug showed a negative relationship.
“At first, that looks strange,” Zhao said. “But once you think about seasonal health patterns, it starts to make sense.”
Cold weather can increase cardiovascular risks, while respiratory issues are often more common in warmer or drier conditions. Understanding patterns like these helps retailers plan inventory, staffing, and promotions more effectively.
For business leaders, the takeaway is clear. Better data and smarter models lead to better decisions.
“Sales forecasting isn’t just about numbers,” Zhao said. “It’s about understanding customer behavior in context.”
That insight is increasingly important as companies face faster product cycles, tighter margins, and more complex supply chains. Zhao’s research shows that looking beyond individual products can give organizations a clearer view of what’s coming next, and a competitive edge in preparing for it.
| Research@Lubar Faculty scholarship in the Lubar College of Business spans the business fields and beyond through both theoretical and applied research that is published in leading journals. Here are some of our faculty’s most recent publications: |
Marketing Science Author: Zuhui Xiao |
Applied Economics Authors: Steve Michael and Antonio C. J. Porto |
Journal of Operations Management Authors: Onkar S. Malgonde, Moez Farokhnia Hamedani, Sunil Mithas, Manish Agrawal, Kaushal Chari |
| Click here to see more faculty research |
Information Systems Research
Authors: Jing Liu, Gang Wang, Huimin Zhao, Mingfeng Lu, Lihua Huang, and Gang Chen ↩︎