Introduction
When buyers search on eBay, the default order of results is called Best Match. It's designed to show the most relevant listings and incorporates a variety of quality, demand and market-driven factors, such as item popularity, pricing, shipping, regions, seller details, etc. eBay has significantly invested in efforts to enhance the machine learning algorithms to improve the quality of search results for buyers. The most recent update to Best Match personalizes the search results for each buyer.
eBay’s marketplace is extremely diverse. In a single category you can find inventory ranging from a few cents to hundreds or thousands of dollars. There are 174 million active buyers on eBay, and every individual has their own criteria to define a perfect item and price range for the product they are shopping for. Previously, when users searched for a product on eBay, the Best Match algorithm would show the exact same results, irrespective of the individual needs. We introduced the price propensity feature in search ranking to customize the search results based on a user’s price preference.
Price propensity takes into consideration a user's past purchases at eBay, as well as the inventory they were interested in. Based on this information, a machine learned model tries to predict user’s price propensity for future purchases and uses this to surface search results that are personalized to the user’s price preferences. As an outcome, we see customized search results for individual users according to their preferences and higher engagement with those results.
Conclusion
Price propensity is just a first step toward personalization in ranking. There are many other factors that we are exploring as part of this work stream. With these features in place, we hope to make the eBay shopping journey for a buyer a simple and efficient experience.