Delete raw events on a schedule

How to do it

Set the period in the privacy policy and in one setting. Hosted tools have a setting for it, such as Google Analytics 4's choice of 2 or 14 months. For your own store, partition the events table by month and drop old partitions, which is faster than deleting rows. Build daily aggregates without IDs first, so long-term charts keep working. Include experiment exposure logs and the rejects table. The general rule is delete old data automatically.

Example

-- PostgreSQL: events partitioned by month; the job detaches and drops partitions past the limit
ALTER TABLE events DETACH PARTITION events_2025_07;
DROP TABLE events_2025_07;

const RETENTION_MONTHS = 14;
export async function dropExpiredPartitions(now = new Date()) {
  const cutoff = new Date(Date.UTC(now.getUTCFullYear(), now.getUTCMonth() - RETENTION_MONTHS, 1));
  for (const partition of await db.eventPartitionsBefore(cutoff)) await db.dropPartition(partition);
}

from datetime import date

RETENTION_MONTHS = 14

def drop_expired_partitions(today: date):
    months = today.year * 12 + today.month - 1 - RETENTION_MONTHS
    cutoff = date(months // 12, months % 12 + 1, 1)
    for partition in db.event_partitions_before(cutoff):
        db.drop_partition(partition)

How to check

A query finds no raw event older than the stated period. The aggregate table still has rows from before it, and none of them hold an ID.