Data Retention Policy
A defined rule for how long each class of data is kept, where, and what happens at the end of it.
Three forces pull in different directions and a policy is where they get resolved. Regulation may require keeping records for seven years, and may separately require deleting personal data on request. Cost rises with every byte kept in a hot tier forever. Risk rises too: data you no longer hold cannot be breached, so retention is a security control as much as a storage one.
A workable policy is per data class, not global, and names the retention period, the storage tier over time (hot → infrequent → archive), the deletion trigger, and who owns the decision.
Two architectural implications people hit late. Deletion must be implementable — an append-only
event log, an immutable backup, and a downstream analytics copy each make erasure a design problem
rather than a DELETE. And retention must be enforced by an automated lifecycle rule, because a
policy that depends on somebody remembering is not in force.