Backups and portability

Data portability is an architectural commitment, not a checkbox. Every table is documented and exportable in three formats.

Exporting

From the admin dashboard, Settings → Export data. You’ll get a zip containing:

clicks.csv
clicks.json
identities.csv
identities.json
events.csv
events.json
attributions.csv  (derived — re-runnable)
commissions.csv   (ledger — immutable)
payouts.csv       (ledger — immutable)
schema.sql        (full schema)
data.sql          (full pg_dump-compatible dump)

The CSV / JSON layers are stable across versions. The SQL dump is the fastest path for re-importing.

Importing

The CLI accepts a hosted export as input:

openpartner import ./openpartner-export-2026-04-24.zip

This re-creates clicks, identities, and events. Attribution and commissions are re-derived from raw data using the configured model — so if you’ve changed your attribution model in the new instance, the imported history reflects the new model.

Why this matters

Most attribution platforms collapse raw events into a single attribution decision and store only that. When you export, you get a snapshot of decisions — not the data those decisions were made from. You can’t migrate, you can’t re-run, and you can’t audit.

OpenPartner stores the raw layers immutably and treats attribution as a query. Export → import → re-derive is lossless by design.