| dc.description.abstract | With the popularity of data mining technology, companies are urged to invest in big data and try to improve their profits by digging out potential customers and reducing the churn rate. However, e-commerce auction platforms have low entry barriers, which attract a large number of sellers to enter so that both competition and the threat from potential entrants are higher. In order to increase revenues in a more cost-effective way, the seller need not only to focus on the returning customers, but also to reduce the churn rate. This study aims to establish an efficient and effective model to predict the churn of buyers. We define the buyer’s churn as the customer does not return and purchase for a certain period of time in the future. We will predict the churn from the perspectives of both the seller and the platform.
The dataset of this study is collected from Yahoo! Taiwan and Ruten auction websites, including all transactions and products information dated from January 1, 2014 to March 31, 2015. Through these empirical data, we establish a prediction model of the churn in auction platforms. We use descriptive statistics to show the consumer’s behavior and consumer inertia, and use six predictors, including 4 RFM-related and 2 augmented variables to predict the churn of buyers. Results show that the buyers are most likely to churn if they have longer last purchase interval, fewer transactions, lower total transaction amount, lower transaction diversity, and lower loyalty type. Average transaction amount is also significant but in reverse direction for the seller and the platform. Major contributions of this study includes: (1) comparing the consumer behavior of the two largest auction platforms in Taiwan; (2) helping sellers identify existing customers who are likely to churn; (3) the results of the descriptive statistics can be used as a reference for the future studies.
| en_US |