Use case · Medallion, modelled, enforced
A governed lakehouse you operate
Build a bronze–silver–gold lakehouse on open table formats and your own object storage, with contracts enforced between layers, dimensional models as first-class objects, and a SQL endpoint that applies row-level security.
- Storage
- Open table formats, your object store
- Layers
- Bronze · silver · gold, contracts on write
- Access
- Row-level security, column masking
- Status
- Early access — design partners
Who it is for
For data platform teams who need a warehouse they fully control, and who want 'medallion' to mean something enforced rather than a folder naming convention.
What this looks like from the inside
Most 'medallion architectures' are a naming convention everyone agreed to once. Nothing enforces the contract between layers, 'customer' means three different things in three reports, and the access rules live in whichever client happened to query.
A warehouse you rent solves some of that and creates a new problem: the business's numbers now live on someone else's cluster.
Drawn, not described
The one thing the prose cannot do: show the moving parts and where the boundary is.
governed SQL endpoint
one door for analysts, Studio and assistants · identity-scoped · audited
gold
facts + conformed dimensions
business-ready
silver
conformed, typed, deduplicated
contract checked on write
bronze
raw, as it arrived
landed by Flow, run recorded
A breaking upstream change fails at the gate, loudly — not in a report, quietly.
Step by step
In the order it actually happens.
- 01
Land raw data in bronze, as it arrived
Revoluzen Flow pipelines write source extracts into the bronze layer in open table formats on your object storage, unchanged, with the run recorded step by step.
- 02
Promote to silver through a contract
Conformance into silver is a recorded operation. The contract — schema, types, expectations — is checked on write; a breaking upstream change fails loudly here instead of silently corrupting gold.
- 03
Model gold around conformed dimensions
Facts and dimensions are catalog objects. 'Customer' is defined once and shared across subject areas — and with Revoluzen Studio — so a pipeline, a query and a report agree by construction.
- 04
Serve through the governed SQL endpoint
Analysts and tools query one endpoint. Row-level security and column masking are enforced there, scoped to the identity asking; the query is audited.
- 05
Change with lineage, not guesswork
Column-level lineage from source to report, with ownership on every dataset. Impact analysis before a schema change — the blast radius is a query, not an incident.
What actually changes
| Dimension | Today | With Revoluzen |
|---|---|---|
| What 'medallion' means | Three folder prefixes and a slide. | Three layers the platform knows about, with contracts checked on write and promotion recorded. |
| What 'customer' means | Whatever each report's author decided that week. | One conformed dimension, shared by pipelines, queries and reports. |
| Where access rules live | In each client, and in the hope that nobody connects directly. | At the endpoint, enforced server-side, audited. |
Which product does what
Each product is adoptable on its own; together they share one catalog, one permission model and one run history.
Revoluzen Catalog
Medallion layers with contracts checked on write; facts, dimensions and conformed dimensions as catalog objects; column-level lineage; a governed SQL endpoint with row-level security and column masking.
Product pageRevoluzen Flow
Feeds the layers with pipelines whose runs are recorded step by step and can be signed.
Product pageRevoluzen Studio
Reports against the same conformed dimensions, so a dashboard and a filing agree by construction.
Product pageWhere it runs
Your object storage, your compute, open table formats. Early access today: design-partner deployments with our engineers on the call.
What you end up with
- 01Contracts between layers that fail loudly on write, not quietly in a report
- 02One definition of each dimension, shared by pipelines, queries and reports
- 03Access rules enforced at the endpoint, not trusted to each client
- 04Impact analysis before a schema change, from column-level lineage
The honest limits
You would find these out eventually. Finding them out now is cheaper for both of us.
Revoluzen Catalog is in early access. The architecture is settled and the core works; it does not have a decade of production deployments behind it.
Scale characteristics are being established with design partners. We will talk about your volumes rather than quote a number we have not measured.
It is a lakehouse you operate: it assumes you have, or want, object storage and someone to run it.
Asked about this scenario
What is Revoluzen Catalog?
Revoluzen Catalog is a self-hosted governed lakehouse on open table formats: enforced medallion layers with data contracts, dimensional and conformed-dimension modelling, column-level lineage, and a governed SQL endpoint with row-level security and column masking. It is in early access.
Does Revoluzen Catalog lock my data into a proprietary format?
No. Revoluzen Catalog stores data in open table formats on your own object storage, so the data remains readable by other tools and remains yours if you stop using the product.
How does Revoluzen enforce a medallion architecture?
Bronze, silver and gold are structures the platform knows about. Contracts between layers are checked on write, promotion between layers is a recorded operation, and each layer's schema and expectations are versioned.
Adjacent scenarios
Dashboards without the export
On-premises analytics with governed answers
Dashboards, paginated reports and a governed query layer that run inside your network against your own warehouse — where every number on a screen can say which metric and which query produced it.
Move off per-row pricing
Replace hosted ELT with self-hosted pipelines
Move pipelines off a hosted ELT service onto an orchestrator you run yourself — with a visual canvas, SQL and Python steps, resumable runs and per-environment licensing that does not grow with your row count.
Run it yourself, then decide.
Revoluzen Flow is Apache-2.0 and installs in one command. If it fits, we can talk about the parts that are not open source.