Designing a production search service for an e-commerce website.

摘要

It is very important for an e-commerce website to provide an effective on-site search service to help visitors find what they want. A higher successful search rate (SSR) leads to a higher conversion rate (CR). This thesis is trying to answer the question of how does a specialized e-commerce website develop suitable on-site search solutions. This thesis introduces general approaches to improve the SSR of e-commerce websites. The production websites that are being studied are very specialized, they are part ordering systems containing several million types of parts. Many of the general design rules will not be applicable. This thesis introduces specialized improvements to search processes. Fuzzy Search and Auto Complete will be focused on in particular. Two experiment plans were implemented and deployed in the production system. After analyzing the website data, fuzzy search showed statistically significant improvement in increasing the SSR and CR while auto complete showed less improvement. A test application was also designed and implemented to compare two of the most popular text search solutions. The testing result shows Lucene[24] is a more efficient and powerful solution than full text search in MS SQL 2005. Both fuzzy search and auto complete solutions in this thesis were implemented using Lucene and JQuery.