We presently have Solr deployed across multiple servers on our image sharing site. We

Question

0

Asked: May 27, 20262026-05-27T15:29:53+00:00 2026-05-27T15:29:53+00:00

We presently have Solr deployed across multiple servers on our image sharing site. We

0

We presently have Solr deployed across multiple servers on our image sharing site. We have 10 million images, with 1/4 million added monthly.

So far Solr does a very good job at selecting search results, but we think there is room to improve on sorting/ranking. We think that incorporating click-through rates in the rankings would provide significantly improved results.

We presently collect click-through data via MongoDB. We record how many times an image is clicked on vs. how many times it is displayed, per term. So for example:

[image identifier], [search term], [click-through rate]

"00000001", "banana peel", "0.1565"

"00000001", "banana", "0.0216"

"00000001", "monkey banana", "0.0087"

What we want to do is find a way to incorporate this search-term-specific click-through data into our Solr rankings. The more an image has been clicked on for that same term, the higher it will rank. We have yet been able to find a way to do this cleanly.

We would like to load the data externally, and not have it as part of the Solr index – as we’d like the click-through data to be nearly real-time and would like to keep our Solr catalogue from getting too huge.

Any ideas or thoughts would be very much appreciated!