Predicting eBay Listing Conversion

Ted Yuan, Zhaohui Chen, eBay

Abstract

Online auction listing conversion (item sold) rate is a statistical value and measured by number of items sold divided by number of items available in a given period. Listing item conversion rate can also be expressed as probability of sell for a given item. By investigating eBay listing statistical behaviors, as well as item attributes and click-through data, we developed an online market place listing conversion model that allows us to predict a live listing’s conversion probability with a computed score. Based on statistical analysis and case studies, we have developed heuristic models that consist of supply and demand components. We enhanced prediction result by using machine learning techniques, such as decision trees and logistic regression, and developed listing sell prediction models with good accuracy across categories and sell formats. To apply the sell prediction models in ranking and other work flows, we developed two categories of sell prediction models, one to predict listing sell probability at listing time without click and impression, and another to include runtime item click and impression data. In the ranking case, we believe listings should be ranked higher if they are more likely to sell giving everything else equal. Such sell prediction is also valuable in analysis of inventory quality, runtime filtering, search engine optimization. The work reveals uniqueness and similarity between web search and sells oriented search at eBay.

Bio

Ted Yuan is a software engineer at eBay Inc.. He has been working on search ranking related projects, especially listing classification design and development, in the Search Science Factors team since joined eBay in 2008. Prior to eBay, Ted worked as individual contributor for Yahoo! in its desktop search and toolbar teams. Before Yahoo!, Ted worked in many start-ups, including social search engine Spoke Software, web performance measurement and monitoring firm   Keynote Systems. Ted holds a doctorate in physics from Northeastern University in Boston and a bachelor's in high energy physics from University of Science and Technology of China.

Zhaohui Chen is now Director of Search Science Factors team at eBay Inc. He has been working on web search, ecommerce search, data warehouse and business intelligence areas. Before joining eBay, he was a Senior Director of Business Intelligence at Taobao.com. Before Taobao.com, he was a senior software engineer at Yahoo! Search Technology and worked on web page classifiers and search engine metrics. Zhaohui holds a PhD in Engineering Science from Harvard University and a bachelor's in Electrical Engineering from Xi'an Jiaotong University, China.