Last I checked, RethinkDB had poor aggregation performance — which is key to doing analytics, unless you want to use it purely as a master data store and do aggregations elsewhere.

Some simple "group by" selects that would take ~3 seconds with Postgres or Elasticsearch would take several minutes with RethinkDB, unless it died of RAM starvage first. It looks to me like RethinkDB is not optimized for sequential read access, nor is its caching algorithm tuned to such workloads.

I believe it also lacks many of the aggregations that you'll want to use, like multi-level bucketing on different dimensions.

This was ~1 year ago, though, so it may have massive improved since then, who knows.