Recovery Point Objective
also called RPO
The maximum acceptable data loss measured as a duration, determining the replication and backup strategy.
RPO = 0 requires synchronous replication, which means every write waits for a remote acknowledgement. Across a region that is tens of milliseconds added to every write, and it makes the remote site a write-path dependency: if it is unreachable, you either stop accepting writes or abandon the guarantee.
RPO in seconds or minutes allows asynchronous replication — no latency cost, at the price of losing whatever was in flight.
RPO in hours allows periodic backups, which is what most non-critical systems actually need.
The distinctions that matter in practice: replication is not backup. Replication faithfully copies corruption, deletion and a bad migration to the replica in milliseconds. Only backups with retention protect against logical damage, and point-in-time recovery is what turns "we have backups" into a real RPO.
And the number is only real if restores are tested. Backup success metrics measure the backup job, not recoverability — a distinction that has ended companies.