Data Dictionary
Human-readable definitions for every table and field — what it means, not just what type it is.
High-Level Design
The Data Dictionary adds meaning on top of the technical schema.
💼 Business Context
- A schema tells you a field's type; a dictionary tells you what it means and how to use it correctly — both are necessary for self-service analytics
- Reduces the "what does this column actually mean" questions that otherwise land on data engineering
- Owned by Analytics Engineering, with definitions contributed by each business domain
🔌 Technical Overview
The data dictionary layers business-readable descriptions, valid value lists, and usage notes on top of the technical schema stored in the Catalog, managed through Azure Purview's business glossary feature. Fields can be linked to a canonical business term (e.g., 'Customer Lifetime Value') with a single definition shared across every table that includes that concept, rather than each table's schema comment drifting independently.
Dictionary Contents
💾 Dictionary Entry
{
"field": "marts.customer_ltv.ltv_band",
"business_term": "Customer Lifetime Value Band",
"description": "Tier bucket derived from trailing-12-month LTV",
"valid_values": ["bronze", "silver", "gold", "platinum"]
}
🔗 Integration Points
- Azure Purview business glossary — the dictionary implementation
- Catalog — technical schema that the dictionary adds meaning on top of
- Query & Analytics Engine's semantic layer — surfaces definitions in BI tools
- Domain teams — contribute and own definitions for their business terms
🧰 Services Consumed
- Owning microservice —
Cxos.Foundation.Api(see the Full Application Service Map) - Database — ADLS Gen2 (Iceberg) + Azure Database for PostgreSQL (policy/retention state)
⚠️ Non-Functional Considerations
- Scale: dictionary entries grow with business term count, not data volume — a curation effort, not an infrastructure concern
- Latency: not applicable — this is reference metadata, not a runtime dependency
- Reliability: canonical term linking prevents definition drift across tables that reuse the same concept
- Security/Privacy: dictionary entries can flag a field as PII-adjacent even before formal classification is applied
🎯 Enterprise Example
A marketing analyst building a self-service dashboard hovers over 'ltv_band' and sees its business definition and valid values directly in the BI tool, instead of pinging data engineering to ask what the field means.