Skip to main content
Datenschaftler

Solution pattern · reference architecture

Lakehouse foundation for the Mittelstand.

This pattern creates a controlled data foundation for reporting, analytics, and later AI applications. Bronze, Silver, and Gold separate unchanged inputs, quality-assured data, and business-approved data products.

Starting point

Reports depend on exports, isolated solutions, and conflicting data snapshots.

ERP, CRM, production systems, and files use different terms, refresh cycles, and ownership models. New analysis becomes another integration project while lineage and quality remain difficult to explain.

Business value

A shared data foundation makes decisions traceable.

Sources remain available in their raw state, cleansing is versioned, and approved data products have clear owners. Reporting and analytics can use the same definitions.

What becomes measurable

  • Freshness and completeness by source and data product
  • Errors, quarantined records, and manual corrections at each processing step
  • Usage, lineage, and business approval of Gold data products

Reference architecture

Components and their purpose

The layers separate ingestion, quality, business logic, and consumption. Azure Data Lake Storage holds the data, Databricks processes it, and Unity Catalog controls access and lineage.

  1. 01

    Connect sources

    Azure Data Factory or existing interfaces ingest selected data from ERP, CRM, files, or databases. Load time and source version are logged.

  2. 02

    Secure the Bronze layer

    Azure Data Lake Storage Gen2 and Delta Lake retain ingested data as close to its original state as possible. Repeatable runs and traceability remain available.

  3. 03

    Validate the Silver layer

    Azure Databricks standardises types, keys, and master data. Invalid records are visibly separated instead of being silently discarded.

  4. 04

    Model the Gold layer

    Business-approved measures and data products are versioned and published for finance, sales, or production.

  5. 05

    Provide controlled consumption

    Power BI, Databricks SQL, or approved applications use defined Gold models rather than unvalidated raw data.

  6. 06

    Govern access and operations

    Unity Catalog, Microsoft Entra ID, Azure Key Vault, and Azure Monitor support roles, secrets, lineage, cost visibility, and operational monitoring.

Technology

Concrete services for implementation

The selection is adapted to existing contracts, regions, security requirements, and the actual scope.

  • Azure Data Factory
  • Azure Data Lake Storage Gen2
  • Azure Databricks
  • Delta Lake
  • Unity Catalog
  • Azure Key Vault
  • Azure Monitor
  • Power BI

First project scope

A pilot needs clear boundaries

The first deployment tests data, integration, and the working process in a limited area. It is not a premature enterprise rollout.

Deliberately included

One relevant data product with a small set of named sources, agreed quality rules, an accountable business owner, and an existing report as the acceptance context.

Deliberately excluded

No enterprise-wide data migration, replacement of every existing report, real-time requirement, or AI application before data quality and accountability have been established.

Prerequisites and constraints

Technology does not replace data accountability

Data access, responsibilities, licences, and operations must be clear before implementation. Open points are treated as project risks.

  1. 01Source access, extraction windows, and network paths must be agreed with the respective system owners.
  2. 02Keys, accounts, customers, materials, or assets need business-maintained mappings. A lakehouse does not repair missing master data automatically.
  3. 03Databricks, Azure, and Power BI licences and the expected storage, network, and compute costs must suit the operating model.
  4. 04The work needs a business data owner, source-system knowledge, platform operations, and input from data protection and information security.

Germany and the EU

Compliance follows the specific purpose

Personal data is subject to GDPR requirements for purpose limitation, minimisation, access, and deletion. EU regions and private network paths should be selected where the required services are available, and actual processing paths must be verified contractually. If data can reveal information about employees, the works council must be involved early under the Works Constitution Act. The EU AI Act becomes relevant when an AI system with a corresponding intended purpose is operated on this foundation.

Free initial assessment

Which data product could replace the most conflicting exports today?

The discovery call can frame sources, definitions, accountability, and a deliberately bounded pilot.

The initial consultation and joint use-case discovery are free and non-binding.