[rchasman]

We use ElasticSearch at Lawmatics, and it powers more functionality than just our search! We use it to power our Automation targeting engine, reporting features, audience builder and pagination, filtering, sorting of data tables.

We denormalize associated records into one Index. And any record that we need to find based on user-defined queries will go through ES since it's much simpler to metaprogram queries across denormalized data (no conditional joins).

[yehosef]

This is one of the undersung benefits of ES in my eyes. Relevancy results requires tuning of the indices and queries and in most cases (that non IR-experts would program) ES will give as good results and be as easy or easier to implement as Solr.

But after you've gotten over that, you realize that this new tool can do lots more things than just text search. Time series metrics, BI, predictive ML, APM, etc. with relatively little work. With Solr, you could do those non-IR tasks, but it's going to feel much more awkward, IMO.