Modeling method provides smarter way to predict customer demand

Published in the journal Production and Operations Management, the study introduces a mathematical modeling method that enables businesses to estimate customer interest beyond just completed transactions and traditional forecasting techniques. The approach offers a more precise way to understand demand, optimize operations and improve decision-making.

"Most businesses can only see part of the demand picture -- they know who buys but not how many people considered buying and didn't," said lead author Xinchang Wang, an assistant professor of operations management at WSU's Carson College of Business. "Our model reconstructs the missing pieces, giving companies a more complete and reliable demand estimate."

Businesses in industries such as travel, hospitality, retail and e-commerce have long struggled with accurately forecasting demand. Many rely on broad assumptions, such as estimating total market size based on their market share. According to Wang, these traditional methods often fail to capture actual customer behavior, leading to inaccurate sales projections and missed revenue opportunities.

Wang and his co-author Weikun Xu, a Carson PhD student in management science, developed a new approach that estimates not just sales but the total number of customers considering a purchase. By analyzing real-world sales data more accurately, the model provides a clearer view of how many customers walked away due to pricing, timing, or other factors.

To develop their model, the researchers used a computational technique called the sequential minorization-maximization algorithm, which improves demand forecasting accuracy. Unlike traditional methods, which can generate multiple possible demand estimates with no clear way to determine the best one, this algorithm -- under specific data conditions identified in their study -- ensures a single, most accurate prediction. "By eliminating uncertainty, businesses can make more confident pricing decisions," Wang said.

Because the model was developed to work with incomplete data, its applications extend beyond any single industry.

While the study tested the model using airline ticket sales data, Wang said the method is designed to be applied across industries where businesses face similar demand uncertainty.

Hotels could use it to predict bookings even when travelers browse but don't make reservations. Retailers and grocers could apply it to estimate total market demand, even when some customers shop at competitors. E-commerce platforms could better understand shopping cart abandonment and refine sales strategies accordingly.

"This model provides a powerful tool for industries where incomplete data has been a persistent challenge," Wang said. "By improving demand forecasting, businesses can plan more effectively, optimize operations and ultimately become more competitive."