Master data management in Microsoft Fabric with Profisee

Organizations adopting Microsoft Fabric quickly discover that a unified analytics platform does not unify the data itself. If the CRM, ERP and e-commerce systems hold three versions of a customer, OneLake will hold all three. This guide explains how master data management fits into a Fabric architecture, how Profisee and Microsoft Purview divide the work, and what to plan for. Woven Measure is a Profisee Certified Partner (Consultant/System Integrator); facts about Microsoft and Profisee capabilities below are linked to their published documentation. Check current documentation before you make design decisions, because both products change frequently.

Why Fabric needs master data management

Fabric brings data engineering, warehousing, real-time analytics and Power BI onto one platform with OneLake as shared storage. It solves the problem of where data lives and how teams work with it. It does not decide which customer record is correct, which supplier is a duplicate or which product description is authoritative. Without that, every semantic model, report and AI agent has to handle conflicts itself, and usually handles them differently.

Master data management adds a governed layer that produces one golden record per entity, with survivorship rules, stewardship workflows and an audit trail. In a medallion architecture (bronze, silver, gold), mastered data becomes the trusted, conformed dimension data used in the gold layer.

What Profisee provides in Fabric

According to Profisee, it offers a generally available, native master data management workload in Microsoft Fabric (Profisee: Microsoft Fabric). Profisee's September 2025 announcement describes capabilities that let Fabric users:

• Connect OneLake datasets to Profisee entities for bi-directional integration, with mappings and transformation expressions configured inside Fabric.
• Define, run and monitor Profisee data quality rules from within the Fabric interface.
• Publish golden records to OneLake for downstream analytics.

(Profisee press release, Sept. 16, 2025.)

How Microsoft Purview and Profisee divide the work

Microsoft's reference architecture for Purview and Profisee (Microsoft Learn) describes a clear split:

In short: Purview tells you what data you have and how it should be governed; Profisee changes the records so they meet those rules.

Reference architecture: master data in a Fabric estate

1. Ingest: source data (ERP, CRM, e-commerce, supply chain) lands in OneLake (bronze) via Fabric Data Factory, mirroring or existing pipelines.
2. Catalog: Purview scans sources and OneLake; business owners define glossary terms and classifications for master data domains.
3. Master: Profisee reads source records (through its Fabric workload integration, Azure Data Factory or APIs), standardizes them, matches and merges duplicates, applies survivorship and routes exceptions to stewards.
4. Publish: golden records and cross-reference keys are published back to OneLake as conformed dimensions (silver/gold), and optionally synchronized to operational systems.
5. Consume: Power BI semantic models, data science notebooks, Copilot and AI agents use mastered dimensions. Microsoft documents a Profisee Power Query connector, and Profisee provides a Power BI connector that respects role-based security.

Azure or Fabric: where should the MDM workload run?

Profisee's own guidance (Profisee: Azure) is that most enterprises use both. Run the workload in Fabric if your data and analytics teams primarily work there; use Azure if your estate extends beyond Fabric or you want more control over deployment and integration with Purview, Azure Data Factory and Synapse. Both run the same platform and data model. Practical questions we work through with clients:

• Where do the consuming teams work today, and where will they work in two years?
• Do golden records need to flow back to operational systems (coexistence) or only to analytics (consolidation)?
• What are your network, identity and data-residency requirements?
• How will capacity, licensing and cost be allocated across Fabric and Azure? Fabric is licensed by capacity (see Microsoft Fabric licenses), and Profisee licensing is contracted separately with Profisee.

Implementation steps

1. Pick the domain and outcome. Customer and product dimensions are common first targets because they feed most Power BI models.
2. Align with governance. Agree on glossary terms and owners in Purview before modeling.
3. Model and map. Build the Profisee model; map OneLake source tables to entities.
4. Standardize, match and tune. Use real data and steward reviews to tune thresholds. See our implementation roadmap.
5. Publish and repoint. Publish golden records to OneLake and repoint semantic models to the mastered dimensions.
6. Measure. Track duplicates removed, completeness and report reconciliation issues before and after.

Common pitfalls

• Mastering in notebooks. Ad-hoc deduplication in Spark or SQL is hard to govern, audit and maintain. Use a governed hub for anything business-critical.
• Skipping stewardship. Automated matching still produces exceptions; someone has to own them.
• Analytics-only thinking. If operational systems keep creating duplicates, the clean-up never ends. Plan the path back to sources.
• No cross-reference keys. Publish source-to-golden key maps so facts from every system can join to the mastered dimension.

Frequently asked questions

Related guides and services

Related Woven Measure products

Check-It runs scored assessments and audits; we use Check-It internally. Praxis Point brings structure, lineage and governance to your data ecosystem. Secur-A-Suite provides encrypted collaboration spaces for sensitive project work.