Glossary
The vocabulary of data projects, without needless jargon.
Much of the misunderstanding in a data project comes from words used differently by technical teams and business teams. Here are the definitions we use, written to be understood without prior training.
Definitions
The recurring terms
- Business Key
- The identifier an organization uses to designate a real-world object: a customer number, a contract reference, an employee ID. It's used to match records coming from different systems.
- Data Vault 2.0.
- A modeling approach that separates identifiers, the links between them, and descriptive attributes. It's designed to absorb new sources without reworking the existing model.
- Reference data
- Data shared across multiple processes — customers, products, organizations — whose definition must be consistent for figures to reconcile.
- Golden record
- The record chosen as authoritative for an entity, built by reconciling multiple sources according to explicit rules.
- Historization
- Keeping the successive states of a piece of data instead of overwriting the previous value. It allows a past situation to be reconstructed and a discrepancy to be explained.
- Information Mart
- An output built for a specific use: a dashboard, a regulatory filing, an application. It's produced from the central model and documented.
- Ingestion
- The operation that pulls data from a source system into the platform, either in full or incrementally.
- Lignage
- The path a piece of data takes, from its source system to the consumed output, including the transformations applied along the way.
- Mapping
- The correspondence between a source's fields and the model's elements. In beVault, it's described, versioned metadata, not code scattered across the codebase.
- Metadata
- The information that describes the data: structure, origin, definition, applied rules. This is what makes it possible to generate the processing and produce the documentation.
- Orchestration
- The coordination of processes: in what order, under what conditions, with what dependencies, and what to do when one of them fails.
- Data Quality
- The set of controls verifying that data is complete, consistent and compliant with business rules, with explicit handling of exceptions.
- Reconciliation
- The check that a figure produced by the platform matches the one in the source system, with an explanation of any discrepancies found.
- Reversibility
- An organization's ability to take back or move its platform without depending indefinitely on a vendor.
- Source of truth
- The reference that has authority over a given piece of data. Without it, every team produces its own figures and no one can arbitrate.
- Data sovereignty
- Control over where data is stored and processed, and who can access it — a strong requirement in the public sector and regulated industries.
Next step