147-2008: A SAS® Text Mining Approach to Predicting the Resolvability of Disputes between eBay’s Sellers and Buyers

A well functioning reputation and feedback system is foundational in consumer-to-consumer electronic commerce, an example of which is the electronic auction. Our interest in this paper is the analysis of the predictive power of both buyer and seller comments in determining the resolvability of transaction disputes in online auctions. Using data gathered from the eBay, Inc. reputation system, we analyze buyer and seller comments using the versatile SAS® Enterprise MinerTM software Text Miner node. In our analysis we employ a binary target variable “Resolvable” to indicate whether an auction dispute has the possibility of being resolved to the satisfaction of buyer and seller. The results suggest that textual analysis of seller comments is consistently more predictive than either content analysis based on human coding of feedback text or textual analysis of buyer feedback alone. The implications and impact of this exploratory analysis are important for potential buyers and sellers in online auctions. First, if a buyer is interested in determining whether a dispute will be resolvable, he/she should spend the most time analyzing the comments from the seller, as the seller seems to have the most power over the manner in which a dispute is resolved, indeed, if it will be resolved at all. Second, when given the choice of examining the final auction price only, or the final auction price and seller comments, a potential buyer should opt for examining seller comments because they are much more predictive of the final outcome of the dispute.