When making online purchases, consumers often encounter a lengthy list of products whose quality they know little or nothing about, ranked by an unknown algorithm. Users often turn to others’ online reviews to estimate product quality—a process known as social learning. HKUST’s Dongwook Shin and colleagues shed light on the challenges that platforms face when social learning requires consumers to choose from multiple products. Their study will help platforms choose the most remunerative product ranking strategy in today’s world of competitive product choice.
On most online review platforms, multiple products are displayed with a ranking, requiring consumers to exert cognitive (and physical) effort while scrolling down a list. This exertion can be modeled as a “search cost.” “Because of this extra cost,” say the researchers, “high-quality products may remain underexplored relative to other (possibly inferior) products.” This factor can cost platforms money in the long run.
However, this ranking effect can also offer platforms a “control mechanism”. By adjusting product ranking, a platform can influence social learning dynamics to ultimately maximize the revenue collected from commission on sold items. “Without an understanding of consumer learning dynamics and the influence of product choice,” the researchers note, “the question of how to best choose the product ranking cannot be properly addressed.” Their goal was to provide such an understanding.
To compare the ranking strategies available to platforms, the study used large-market (fluid) approximation, a mathematical technique designed to model markets with many participants. “We study a model of a marketplace where consumers arrive sequentially over time and decide whether to buy one of the available products or to take an outside option,” the researchers say. The probability of a purchase taking place “depends on consumers’ quality estimates, their idiosyncratic preferences, and the prices of the available products.”
A platform can influence the information that potential buyers obtain by altering the order in which products are displayed. “By picking the product ranking, the platform can affect product choice and stimulate information acquisition to ultimately maximize revenues,” the authors write. Information––that is, reviews––accumulates faster for higher-ranked products, “as these products will be selected more frequently and produce more reviews.”
The authors explored how platforms can “judiciously” balance the trade-off between exploration (acquiring information about product quality) and exploitation (optimizing rankings and revenue). They found that a “greedy” ranking policy, which at any given moment ranked products to maximize revenue in the next, was “underexplorative.” To tackle this problem, the authors analyzed a “semigreedy” policy that ranked products based on a combination of their estimated quality and the extent to which they had been explored. Surprisingly, in different settings with different search costs, regret grew at a slower pace than under the “greedy” policy.
These findings offer “qualitative insights into the benefit of exploration relative to the extent to which the ranking affects consumers’ decision making,” the authors write. Thanks to their study, platforms now have a reference for choosing the right product ranking strategy in today’s highly competitive e-commerce industry.