It's the measurement of the amount of data that can be acceptably lost (in seconds, minutes or hours)

of course nobody wants to lose data, but the less data we accept to lose the higher the costs for maintenance and backups.

If we could afford losing 2 hours of Data in a Database , we can configure backups every 2 hours. If we want to reduce the data loss, we need to do more frequent backups.

The lower the RPO the higher the costs and the complexity of the infrastructure ( for example, in case of minutes to hours, snapshots are ok, for seconds to minutes, async replication would be fine, but to get to milliseconds and seconds we would need synchronous replication).