HyperLogLog, also by citusdata (no affiliation), is a great data type for statistically approximating counts of distinct values very quickly within an adjustabl

HyperLogLog, also by citusdata (no affiliation), is a great data type for statistically approximating counts of distinct values very quickly within an adjustable range of error: https://github.com/citusdata/postgresql-hll https://github.com/citusdata/postgresql-hll This has been a great secret weapon for me in a few big counting problems.