Schema as data: what a no-code CMS actually costs

One backend serving many client websites, each defining its own content shapes from the dashboard. A client adds a testimonials collection with author, quote and rating, and starts storing rows. No migration, no deploy, no engineer.

That is a genuinely good product property, and it is worth being precise about the cost, because every no-code builder makes this trade and not all of them admit it.

Content fields cannot be indexed, because field names do not exist when the index would be created. Any filter on collection content is a scan. Declared field types are not enforced by the database, so validation has to be rebuilt at the application layer rather than inherited. Renaming a field orphans data stored under the old key unless you migrate it yourself.

None of that argues against the design. It argues for knowing which guarantees you handed back to yourself when the schema moved out of the database and into a document, and rebuilding them on purpose instead of discovering their absence in a bug report.