Understanding the Dynamics of Electronic Word of Mouth.

摘要

This dissertation studies electronic Word of Mouth in the form of consumer reviews. Today, consumer reviews have become an integral and influential part of a customer's purchase journey, prompting marketing professionals and academic researchers to study and try to understand them. Our work contributes both in the literature that tries to understand the effect that consumer reviews have on the customers' purchase decision as well as in the literature trying to better understand and mitigate the biases that may emerge in the current practices of collecting, aggregating and displaying reviews.;First, we study consumer reviews in the mobile app market, a highly valued and competitive market in which developers struggle to make their apps successful. We collect and study a large dataset of reviews from Apple's iOS AppStore and try to understand the correlations between ratings and version updates. Motivated by current policies and by what we see in the data, we formulate and study a mathematical model and find that, within our model, the optimal version update strategy for the developers is a threshold policy on the ratings the version receives. We furthermore demonstrate the explaining power of our model by fitting it to the collected dataset and using it to estimate the quality of the versions released in the store. We compare the estimations of our model against the observed ratings that versions received within the first few days of their lifetime, and find that our model performs better than a baseline random-guess model as well as a model in which users are primarily influenced by how new the version is.;Next, but under the same umbrella of trying to understand the influence of consumer reviews on customers, we study how the displayed volume of reviews can affect the purchase propensity for a product. By analyzing a large detailed dataset consisting of every product page view, review page view, sale and submitted review that occurred throughout 2015 for a major novelty items retailer, we estimate that as a product aggregates more reviews it's conversion rate can increase up to three times, amongst the users that display reviews. Our analysis includes fitting an exponential learning curve into the evolution of the purchase propensity as the product receives more reviews, and a control dataset from users that did not display the reviews in order to control for trends that are not related to the increase of the volume of reviews. This finding provides support for theories studying social influence bias on a purchase level, i.e., that users are more likely to buy a product if they feel that other people have bought the product.;Our second line of work tries to understand and mitigate biases in consumer reviews. First, we study selection bias, which is a bias that arises when the sample of users that submit a review for a product is not representative of the entire population of purchasers. We analyze a large dataset of reviews from four major online retailers and find that self-motivated reviews that are submitted through the web are more bi-modular and carry lower ratings than reviews that were submitted as a result of an email prompting. Using the same dataset, we follow a natural experiment approach to understand the effect that the introduction of the email promptings had on the ratings on the platform. We find that email promptings made the ratings of the platform more representative, credible and caused an increase in volume and star-rating.;Collaborating with a review platform provider, we experiment with the display of social signals on the email promptings sent to customers to understand better social influence bias on the evaluation level, i.e., that a customer's opinion about a product will be affected by their peers' opinions' for that product. We find that by announcing to the customers the current state of the reviews, their submitted ratings increase.;Finally, we study promotions, a situation that via decreased prices and/or increased exposition to users can induce complex biases in the purchase decision of customers as well as their subsequent submitted reviews for any purchased products. Previous work has shown that establishments that offer a Groupon see their Yelp ratings suddenly decline.;Our work contributes to this line of work by studying four different promotions offered in the iOS AppStore and how each unique feature of the promotion can affect the sales and ratings of the promoted apps. We find that an organic integration of the coupons in the user experience, as well as careful selection procedure can greatly benefit the sales as well as the ratings of the offered apps. (Abstract shortened by ProQuest.).