Bronze / preserve the source
A faithful starting point.
Capture source data and the metadata needed to understand its origin and loading history. This layer establishes a traceable foundation for downstream integration.
Enterprise architecture best practices
Established patterns.
Proven experience.
The architecture shown here is a common enterprise pattern that our team has successfully implemented. Its principles apply across platforms. We adapt the implementation to your environment, bringing especially deep experience in Azure Synapse and Microsoft Fabric.
From source to shared understanding
Every layer
has a purpose.
This reference architecture combines established medallion, Data Vault, and Kimball dimensional modeling practices with governed semantic models. Our expertise is in implementing these practices together for your business.
-
Sources
Business systems
Operational applications, databases, files, and APIs.
-
Bronze
Source integration
Raw source data, metadata, and captured history.
-
Silver
Enterprise integration
Data Vault business keys, relationships, history, and business rules.
-
Gold
Analytical structure
Kimball facts, dimensions, and conformed dimensions.
-
Semantic
Shared meaning
Governed definitions, measures, relationships, and access rules.
-
Consumers
Many ways to use it
Reports, self-service analysis, and conversational BI with data agents.
Across every layer
- Metadata, catalog & lineage
- Orchestration
- Automation & custom templates
- Governance
- Monitoring
- CI/CD
Silver / integrate the enterprise
Connect identity and history.
Use the Raw Vault to integrate business keys, relationships, and source history. Use the Business Vault to apply reusable business rules while keeping them distinct from source records.
Gold / shape the analysis
Make the data useful to analysts.
Organize information into facts and dimensions at a clearly defined grain. Conformed dimensions give analytical products a consistent way to describe shared business concepts.
Semantic / publish the product
Give the business a common language.
Publish documented, versioned models with shared metrics, KPIs, relationships, and hierarchies. Define row-level and object-level access controls as part of the model design.
Managed self-service analytics
Define once.
Reuse across teams.
A semantic model becomes a reusable product when its definitions, measures, documentation, and access rules are maintained together. Multiple reports can then use the same business logic.
Central teams govern the shared models. Departments build thin reports that connect to them, analysts explore through Excel, and data agents support natural-language questions over the prepared data context.
That separation lets each team shape its view of the business while keeping core definitions consistent.
Explore semantic model deliveryIllustrative implementation / Microsoft stack
Sales semantic model
Agreed measures · relationships · hierarchies · access rules
Power BI
Team reports
Thin reports use shared measures.
Excel
Self-service analysis
Explore the governed model.
Fabric IQ & data agents
Ask the data
Ask questions in business terms.
How the architecture enables AI
Give AI the context
behind the data.
Each layer prepares something AI can reuse. Source capture preserves origin and history. Data Vault connects identities across systems. Dimensional models organize those records for analysis.
Semantic models supply governed measures and calculations. Ontologies describe business entities and their relationships. Data agents use that context to interpret questions, query the right sources, and return answers in business terms.
We have implemented this approach with Fabric IQ, connecting the models and business context to data agents and evaluating responses against agreed business questions.
Explore AI over your business dataA reusable data foundation
Connected data. Preserved history.
- Bronze / source fidelity
- Silver / Data Vault
- Gold / dimensional models
Measures & calculations
Semantic models
Define sales revenue, order counts, and reporting periods.
Entities & relationships
Ontologies
Describe how customers, orders, products, and categories connect.
Illustrative example / Fabric IQ
Fabric data agents
Select the relevant configured source, query it with the user's permissions, and turn results into a response.
How has sales revenue changed by product category and quarter?
Question Select context Query Answer
Designed for delivery
Govern the model.
Give reporting room to move.
Separate model and report lifecycles, with development, staging, and production environments. Promote changes with validation, clear dependencies, and repeatable deployment.
| Delivery track | Development | Staging | Production |
|---|---|---|---|
| Shared models | Build & version | Validate & review | Publish governed products |
| Departmental reports | Create & iterate | Check model connections | Release business views |
Reports connect to the appropriate shared model in each environment. Model definitions and report layouts can evolve through their own release processes.
Automate the standards
Make the patterns repeatable.
WhereScape and custom templates carry architecture conventions into generated models and loading code. Metadata, documentation, and validation help the automation remain maintainable.
Adapt to your platform
Connect design to the platform.
We can implement these patterns on your chosen data platform, adapting storage, integration, and deployment to fit. Azure Synapse and Microsoft Fabric are particular areas of depth for our team.
Evaluate the AI experience
Test answers against the model.
Use representative business questions to check agent responses against shared measures and definitions. Review source access and refine the configuration as models and business needs evolve.
Architecture into practice
Build the foundation
your next use case can reuse.
We can help assess your current platform, define a data product, and put this architecture to work with your team.