Relevance is a hosted site search service, with a user interface for building custom ranking algorithms. Users can A/B test new versions of ranking rules and analyze the results, without needing to code.
As the founder of a marketplace startup, this product made perfect sense to me from day 1. One pain point we have is proper monitoring of when problems arise in the matching process - insufficient supply in a certain segment, bugs introduced by a change that does not seem related but which tweaks the matching algo in some way. Have you thought about building in problem alerts?
@john_gardner Hi John, we plan on expanding our analytics to help monitor search/match effects all the way through the funnel. This can help you determine why search conversion may be slipping in a certain segment, and insufficient supply can definitely be a culprit. I like the idea of problem alerts to be proactive about letting the user know, thanks for the feedback!
@zealoustiger Thanks Brian! You can tweak the importance of different factors in your search algorithm, say average review score vs. number of reviews. Today the tweaking is done by the user (human), but we plan on bringing AI assistance soon (algo) 🙂
Report
Super cool! Is there an upper bound in terms of how many rules you can use in an algorithm?
Replies
Kickoff
Relevance
Weave Product Hunt Group
Relevance
Relevance