Comparative
beVault does not replace Snowflake or Databricks: it gives them a structure.
The question comes up in every selection committee: “we already have Snowflake, why add beVault?” Because those platforms provide storage and compute, not a model. They run the SQL you give them — someone still has to write it, evolve it, schedule it, and check its results. That's exactly beVault's scope.
- Your target platform stays yours: no data is moved
- Native code optimized for each engine
- Modeling, orchestration and quality on top of the compute engine
- Possible outcome: the generated code belongs to you
Division of roles
Two layers, two responsibilities
The confusion comes from the word "platform", used on both sides to mean different things.
Snowflake / Databricks
Storage, compute engine, data access security, elasticity. This is where your tables live and your queries run.
beVault — the model
Hubs, links and satellites compliant with the DV 2.1 standard, with the load code generated for the target engine.
beVault — the execution
Load order, error recovery, and parallelism are all derived from the model's dependency graph.
beVault — trust
Versioned quality rules, per-source scoring, master data reconciliation, lineage and living documentation.
What each brings
Who does what, concretely
| Besoin | beVault | Snowflake / Databricks |
|---|---|---|
| Storage and compute | None: beVault runs on your platform | Core of the product |
| Data Vault modeling | Visual interface, DV 2.1-certified metamodel | Out of scope |
| Load code generation | Automatic, optimized by engine | To be written by your teams |
| Load orchestration | Incluse | Native scheduler or third-party orchestrator to configure |
| Data Quality and MDM | Framework built into the load | Constraints and controls to implement |
| Lineage and documentation | Generated from the metamodel | Platform catalog, populated separately |
| Compute cost control | Incremental loads, no full reloads | Billing based on compute usage |
What changes in practice
The benefit shows over time, not in a demo
A first domain ships quickly either way. The gap appears at the tenth domain, when the number of objects to maintain exceeds what a team can track by hand: that's where using the model as the source of truth changes the project's economics.
Does our data leave our platform?
No. beVault generates and orchestrates the code that runs on your platform. Your data stays within your perimeter.
Is the generated code generic or native?
Native. The performance mechanisms specific to each engine — micro-partitions, distribution, table formats — are leveraged by the generated code.
What if we change target platform?
The model stays the same: generation is redone for the new target, without redoing the modeling.
Next step