News Release 

Recommended for you: Role, impact of tools behind automated product picks explored

Jindal School researchers examine pros, cons of recommender systems

University of Texas at Dallas

Research News


IMAGE: Dr. Jianqing Chen, professor of information systems in the Naveen Jindal School of Management view more 

Credit: UT Dallas

As you scroll through Amazon looking for the perfect product, or flip through titles on Netflix searching for a movie to fit your mood, auto-generated recommendations can help you find exactly what you're looking for among extensive offerings.

These recommender systems are used in retail, entertainment, social networking and more. In a recently published study, two researchers from The University of Texas at Dallas investigated the informative role of these systems and the economic impacts on competing sellers and consumers.

"Recommender systems have become ubiquitous in e-commerce platforms and are touted as sales-support tools that help consumers find their preferred or desired product among the vast variety of products," said Dr. Jianqing Chen, professor of information systems in the Naveen Jindal School of Management. "So far, most of the research has been focused on the technical side of recommender systems, while the research on the economic implications for sellers is limited."

In the study, published in the December 2020 issue of MIS Quarterly, Chen and Dr. Srinivasan Raghunathan, the Ashbel Smith Professor of information systems, developed an analytical model in which sellers sell their products through a common electronic marketplace.

The paper focuses on the informative role of the recommender system: how it affects consumers' decisions by informing them about products about which they otherwise may be unaware. Recommender systems seem attractive to sellers because they do not have to pay the marketplace for receiving recommendations, while traditional advertising is costly.

The researchers note that recommender systems have been reported to increase sales on these marketplaces: More than 35% of what consumers purchase on Amazon and more than 60% of what they watch on Netflix result from recommendations. The systems use information including purchase history, search behavior, demographics and product ratings to predict a user's preferences and recommend the product the consumer is most likely to buy.

While recommender systems introduce consumers to new products and increase the market size -- which benefits sellers -- the free exposure is not necessarily profitable, Chen said.

The researchers found the advertising effect causes sellers to advertise less on their own, and the competition effect causes them to decrease their prices. Sellers also are more likely to benefit from the recommender system only when it has a high precision.

"This means that sellers are likely to benefit from the recommender system only when the recommendations are effective and the products recommended are indeed consumers' preferred products," Chen said.

The researchers determined these results do not change whether sellers use targeted advertising or uniform advertising.

Although the exposure is desirable for sellers, the negative effects on profitability could overshadow the positive effects. Sellers should carefully choose their advertising approach and adopt uniform advertising if they cannot accurately target customers, Chen said.

"Free exposure turns out to not really be free," he said. "To mitigate such a negative effect, sellers should strive to help the marketplace provide effective recommendations. For example, sellers should provide accurate product descriptions, which can help recommender systems provide better matching between products and consumers."

Consumers, on the other hand, benefit both directly and indirectly from recommender systems, Raghunathan said. For example, they might be introduced to a new product or benefit from price competition among sellers.

Conversely, they also might end up paying more than the value of such recommendations in the form of increased prices, Raghunathan said.

"Consumers should embrace recommender systems," he said. "However, sharing additional information, such as their preference in the format of online reviews, with the platform is a double-edged sword. While it can help recommender systems more effectively find a product that a consumer might like, the additional information can be used to increase the recommendation precision, which in turn can reduce the competition pressure on sellers and can be bad for consumers."

The researchers said that although significant efforts are underway to develop more sophisticated recommender systems, the economic implications of these systems are poorly understood.

"The business and societal value of recommender systems cannot be assessed properly unless economic issues surrounding them are examined," Chen said. He and Raghunathan plan to conduct further research on this topic.


Lusi Li PhD'17, now at California State University, Los Angeles, also contributed to the research. The project was part of Li's doctoral dissertation at UT Dallas.

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.