MIT Professor Develops Search Model for
Product Rankings on Online Platforms
Online shopping provides consumers with numerous product options displayed across multiple web pages. As a result, the order in which products are displayed impacts consumer choice. Studying this product-ranking problem, MIT Sloan School of Management Prof. Negin Golrezaei developed a new search model that learns consumer preferences to optimize product rankings for consumers, sellers, and the platform.
“When consumers search for a product, it’s typical for the platform to offer thousands of results, with only a few highlighted at the top of the page. The number of options far exceeds consumers’ attention and cognitive resources,” says Golrezaei. “This creates position bias because those first few products listed will get more visibility even if they aren’t the best matches for the consumers’ needs.”
She notes, “The goal of online shopping platforms is to help consumers find products in the shortest time possible, but if the platform doesn’t do a good job then consumers will feel frustrated or leave without making a purchase.”
In her study, she analyzed a dataset from a platform that displayed a number of options for each search, ranging from a few to a dozen. She determined the probability of purchase for each position for two ranking mechanisms: random ranking and the platform’s optimized ranking. Under both mechanisms, products in lower positions are less likely to be purchased. However, the decline is sharper when the platform’s ranking was used.
“These findings highlight the importance of optimizing ranking in an online platform, but the question then becomes: How should online platforms rank products? What is the optimal method?” says Golrezaei.
She and her colleagues designed a new model of consumer search and choice that captures key features of consumer behavior. The model assumes the utility that a consumer derives from a product consists of two parts: a basic utility and a personal utility.
“The model implements a two-stage process, where in the first stage the consumer screens products for quality and relevance and forms a consideration set. In the second stage, the consumer considers the overall utility that can be derived from the product and chooses the product that best meets their needs in the consideration set,” explains Golrezaei.
Using the model, they studied how platforms with different objectives should rank products. They looked at platforms that seek to maximize the platform’s market share and platforms that seek to maximize consumers’ welfare.
“Somewhat surprisingly, we show that ranking products in decreasing order of consumers’ preferences does not necessarily maximize market share or consumer welfare,” says Golrezaei. “Such a ranking may narrow a customer’s consideration, but the products featured in that consideration set may not meet the individual customer’s needs, resulting in insufficient screening.”
“When consumers run out of patience and don’t buy anything, they are dissatisfied. It’s worth
finding a way to better optimize the products featured in the top spots to ensure consumers
find the best product that matches their unique taste without examining too many products,” she says.
Golrezaei adds, “The bottom line is that product rankings significantly influence demand and platforms can use this knowledge to benefit buyers and sellers as well as the platforms.”