Online product review platforms have become popular places for consumers to share their opinions on and experiences of the products and services they buy. In online marketplaces, the quality of a new product or service is initially unknown, but reviews posted by buyers inform subsequent customers. By reducing uncertainty about product quality, this type of social learning leads to better-informed purchase decisions, say Dongwook Shin of HKUST and two colleagues. Their study, based on data from one of the world’s largest online retailers, offers rich new insights into the interplay between online reviews and pricing policies.
“Product review platforms have emerged as viable mechanisms for sharing opinions and experiences on a wide range of products (and services) in online marketplaces,” say the researchers. “In these markets, the quality of a newly launched product is initially uncertain, but the reviews generated by buyers help inform subsequent customers.” These dynamics, say the researchers, affect the demand trajectory over the selling horizon and present challenges to the seller’s design of revenue-maximizing pricing policies.
To shed light on whether customers can learn a product’s true quality despite the effects of pricing and under what conditions the seller prefers dynamic pricing over fixed pricing, the researchers explored the effects of review information on the seller’s revenue-maximizing pricing decisions and how these decisions affect social learning.
First, the researchers formulated a stylized model of a marketplace in which a revenue-maximizing monopolist sells a product to customers. In their model, they state, “prices affect the dynamics of the review process, and this, in turn, influence the demand and pricing decisions for subsequent customers.”
Next, they conducted a simulation study using data from an online marketplace. “We use data collected from Amazon, one of the largest online retailers in the world,” say the authors. “Our data set includes a thousand hardcover books released between January 2015 and December 2016.” They considered two settings: one featuring quality-based reviews, in which buyers altruistically seek to inform future customers; and another featuring value-based reviews, in which buyers report their perceived utility, which is determined by both the product’s quality and the price paid.
The researchers’ tests yielded a number of important findings. First, they say, “the inherent feedback generated by the presence of reviews significantly complicates” the development of revenue-maximizing policies. They argue that with quality-based reviews, it may be optimal for the seller to use price skimming or penetration pricing, depending on customers’ expectations. With value-based reviews, however, it is always optimal for the seller to use penetration pricing.
In addition, the researchers report, dynamic pricing may be a fruitful way for the seller to respond to the time-varying demand resulting from customers’ review-driven learning process. They find that the benefits of such dynamic pricing depend on both the reviews’ type and their degree of informativeness, which affects the rate of social learning. For quality-based reviews, the researchers say, a fixed-price policy may be near-optimal, because “the learning trajectory, and hence demand, is essentially independent of price.” However, this is not the case for value-based reviews. The researchers conclude that dynamic pricing is most valuable in an environment in which reviews have a moderate degree of informativeness.
On the basis of their findings, the researchers propose a dynamic pricing policy that they call “dynamic fluid matching.” They demonstrate that this policy is asymptotically optimal under quality-based reviews and that it can be computed in closed form. They also develop a fixed-price policy that is near-optimal in situations featuring limited price flexibility. However, the researchers find that for value-based reviews, these are not effective solutions. Instead, “a class of piecewise-constant policies that allocates an appropriate fraction of time to each price” is near-optimal.
In short, by developing models for and analyzing two settings, the researchers comprehensively explore the spectrum of learning dynamics and their ramifications for pricing strategies. Their findings, the authors note, provide “a more comprehensive and nuanced understanding of the interplay between pricing policies and reviews.”