What consumers think before they buy

Manufacturers have traditionally relied on sales figures to understand where they stand with consumers, but by the time those numbers arrive, the consumers' buying decisions have already been made.

A study co-authored by UC Riverside business professor Hai Che offers a way to capture what consumers are thinking while they are still weighing options, comparing products, and deciding whether to buy at all.

The research gleans information from multiple forms of online consumer behaviors, such as text comments that compare car models, to derive what researchers call "competitive intelligence." The approach enables companies to identify competitors, understand consumer perceptions, and predict how new products may fare in the marketplace.

Published in the journal Production and Operations Management, the findings could have far-reaching implications for automakers and manufacturers of many other consumer products because they reveal the thought process consumers go through before making a purchase.

"Sales data tell us what people bought, but not what they were thinking before they bought it," Che said. "Our approach helps automakers understand how consumers perceive their vehicles relative to competitors such as Tesla, and what factors-whether price, features, or brand image-are driving those decisions."

The study focused on the rapidly growing Chinese market for electric and hybrid vehicles, where consumers routinely discuss automobiles on a large third-party online forum called Auto Home. Researchers analyzed how consumers saved vehicles to favorites lists, commented on specific models, and discussed product features in online conversations.

What makes the research significant is its ability to combine these different streams of information into a unified picture of consumer decision-making.

Che describes the data as information "hiding in plain sight." Consumers voluntarily leave clues about their preferences as they browse, save products for later consideration, and participate in online discussions. Individually, those actions provide only partial information. Together, they reveal how consumers perceive competing products and what factors influence their choices.

The researchers developed a framework that integrates three types of user-generated data: favorite lists, products discussed in online forums, and the text of user comments. The model then identifies competitive market segments and the attributes consumers use to compare products within those segments.

The distinction between favorites and comments proved especially important. Products mentioned in discussions often represent early-stage exploration, when consumers are gathering information about options. Favorites lists, by contrast, tend to reflect a later stage in the process when shoppers have narrowed their choices and moved closer to a purchasing decision. By combining both signals, the model captures multiple stages of consumer thinking.

Researchers tested the framework using a massive dataset that included more than 324,000 favorites lists and more than 354,000 user-commented products. Che said the dataset represented thousands of users and more than 2,300 vehicle models.

The results showed the model could more accurately identify competitive submarkets, uncover consumer perceptions, and predict competitive positions for new market entrants. The framework also showed strong alignment with actual sales patterns.

For automakers, those insights could prove invaluable.

"Knowing what people are thinking could be very useful for these automakers," Che said. "They can figure out what's not working so well with a car, whether it could be features, whether it could be consumers' perception of their car relative to a big player."

The broader implications extend well beyond automobiles.

The use of online behavioral data to understand how consumers think during the decision-making process represents a potentially important advance in both consumer psychology and purchasing economics, Che said.

Historically, researchers and companies have relied on surveys, focus groups, and sales records to gauge consumer preferences. The new framework instead extracts insights from naturally occurring online behavior, allowing firms to observe how consumers compare products in real-world settings in real time.

The research grew out of a collaboration that began in 2018 when Che met lead author Yang Qian while giving a research seminar in China. Qian, then a doctoral student with a background in computer science and artificial intelligence, impressed Che with his innovative ideas.

The two began collaborating in 2019, combining Qian's expertise in machine learning with Che's research on consumer behavior and marketing to develop the framework behind the study. Qian is now an associate professor at Hefei University of Technology in China.

The study's title is "Deriving competitive intelligence from multifaceted user behavior data: An interpretable machine learning framework." In addition to Qian and Che, the co-authors are Yezheng Liu, Yuanchun Jiang, and Jennifer Shang.