Office of Technology Transfer – University of Michigan

Method for Ranking Items from Pairwise Preferences

Technology #6514

The proposed technology is a method to approach an extremely prevalent problem—how to provide ranks to some set of items. This happens quite commonly, and is not always simple! For instance, when searching for a partner to date, the best choice isn’t necessarily the most attractive person, but the most attractive person who also finds you sufficiently attractive. This general problem arises in online shopping, dating, ranking of sports teams, search engine queries, and a huge number of other fields. Solving this problem quickly and efficiently provides many different services and benefits!

Outperforming the Best in Ranking Algorithms

Currently, there is no exact algorithm for this kind of problem which can be applied to all cases. Because of this, the best approaches are generally heuristic, and there is always room for improvement in time and accuracy. This technology has been shown to match the accuracy of the best algorithms for this problem, more quickly than the competition.


  • Generating search results from queries in search engines
  • Matching romantic partners online
  • Applications in systems biology to identify most likely drug candidates
  • More accurately targeted marketing for online vendors


  • As accurate as existing methods
  • Significantly faster than existing methods