Backups and continuity

The database is Postgres 17 on Supabase, hosted in AWS eu-west-2 (London, United Kingdom). Uploaded purchase-order files are held in a private storage bucket on the same platform, in the same region.

The complete database schema is defined by migration files held in version control alongside the application source, on GitHub. Continuous integration rebuilds an empty database from those migrations on every change and compares the result, object by object, with production — so the structure can be recreated from source control alone rather than depending on any single running instance.

That is a statement about the structure. It is not a backup of customer data, and we do not describe it as one.

Provider-managed daily backups: yes. Our database provider takes a daily backup with seven days' retention, and we can request a restore to any of them.

Point-in-time recovery: no. It is a paid add-on we have not enabled.

An independent copy held outside the provider: not yet. The tooling to take one exists in the repository; it is not yet running on a schedule.

Uploaded files: not yet backed up separately. Provider database backups exclude file storage by design, so files need their own copy; this is on the same schedule as the database export above.

We have not yet rehearsed a restore. We will not quote a recovery time objective until we have measured one — a number on this page that nobody has tested is a number a reviewer is right to distrust.

The service is provided without a guaranteed uptime level, as the terms say. We will not restate a hosting provider's own durability figures as a commitment of ours.

Every principal record in the product — orders, products, customers, inventory, invoices, reports — can be exported as CSV from its own screen at any time, and remains available for export for the period stated in the terms after termination.