CDM Architecture Overview
Three-Layer Architecture
CDM is implemented across three main data layers, with the Metadata Model as the controlling core:
Layer 1 – Landing Zone (LDZ)- Environment-scoped schema (e.g., dev_ldz, prod_ldz, test_ldz)
- Raw data from source systems, minimal transformations
- Preserves business keys for reconciliation
- Basis for Change Data Capture (CDC)
- Data Vault 2.0 implementation (Hubs, Links, Satellites, Link-Satellites)
- Physical realization of the Canonical Data Model
- Audit-friendly, insert-only, scalable structure
- Central store of canonical entities and relationships
- Customer 360, Campaign 360, and future 360s
- Business-ready, denormalized, analytic views
- Designed for segmentation, analytics, and API consumption
- Focus: "how business users see the world", not how sources store it
Metadata Model – The Core Integration Layer
The Metadata Model orchestrates the full system:
- Integration & mappings: Keeps all mappings from source → LDZ → RDV → 360.
- ETL automation:
Generic ETL engines read metadata and generate:
- LDZ ingestion code
- RDV transformation code
- DQ and lineage code
- Orchestration jobs
- Data quality enforcement: Common rules are defined once and executed everywhere.
- Lineage: Every column, every transformation, end-to-end.
This makes CDM configurable through metadata rather than hard-coded logic.
Metadata Model – The Core Integration Layer Why / What / How for CDM Users
WHY – Problems CDM Solves
- Fragmented customer view across systems
- No consistent customer identity
- Channel-specific silos for campaigns and interactions
- "Next Best Action" downgraded to "educated guess" due to missing context
WHAT – CDM Provides
- Canonical entities (Party, Account, Product, Campaign, Response…)
- Customer 360: unified, behavior-rich, consent- and risk-aware profile
- Campaign 360: standardized performance metrics, audience context, and ROI view
HOW – Implementation Mechanisms
- Canonicalization across products (same structures, different sources).
- End-to-end stitched journeys using identity and events.
- Identity resolution rules embedded in metadata and enforced in RDV and 360.