Master Data Management in practice: styles, tools, and the Golden Record
Ask five data practitioners what Master Data ManagementMaster Data ManagementMaster Data Management (MDM) is the discipline of creating and maintaining a single, consistent, trusted version of an organization's core business entities like customers, products, and suppliers.View full definition → means and you'll get five different answers. Ask them which MDM tool to use and you'll get a 90-minute debate.
MDM is one of the most misunderstood disciplines in data management, and one of the highest-ROI investments a CDO can make when done correctly.
What MDM is (and isn't)
Master Data Management is the discipline of creating and maintaining a single, authoritative, consistently defined version of an organization's most critical data entities, typically customer, product, supplier, and location.
What MDM is NOT:
- It's not a database or a system (though it uses both)
- It's not the same as Data governanceData governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.View full definition → (though governance is essential to MDM)
- It's not a one-time project (it's an ongoing operational capability)
- It's not only about customer data (though that's the most visible use case)
The core deliverable of MDM is the Golden Record: a single, authoritative record for each real-world entity, synthesized from multiple source systems, cleansed, deduplicated, and enriched.
When a retailer has 47 million customer records spread across their CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition →, loyalty program, e-commerce platform, and in-store POS, with 15% duplication, the Golden Record identifies and merges duplicates into 40 million unique, authoritative customer profiles. Every downstream system then works from the Golden Record.
The four MDM implementation styles
Registry style: The MDM hub doesn't store master data, it stores cross-references between source systems. Each source system keeps its own dataown dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.View full definition →; the hub maps "CRM customer #12345" to "loyalty customer #67890" to "billing customer #A4412". Lowest implementation effort, but the master data still lives in silos.
Consolidation style: Source systems remain authoritative, but data is periodically consolidated into the MDM hub for analytics and reporting. Good for read-only use cases; doesn't solve real-time data consistency problems.
Coexistence style: The MDM hub and source systems coexist, with changes synchronized bidirectionally. More complex to implement but gives you both a master view and operational system consistency.
Centralized (hub-and-spoke) style: The MDM hub becomes the system of record. All source systems write to and read from the hub. Maximum data consistency; maximum implementation complexity and organizational change required.
Which style is right for you depends on your integration architecture, organizational appetite for change, and use cases. Most organizations start with Registry and migrate toward Coexistence as maturity grows.
Data Architecture Strategies: Master Data Management
Knowledge check
1. What best describes the 'Golden Record' as the core deliverable of MDM?
2. A company needs both a master view of its data AND real-time consistency in its operational systems, and is willing to accept more integration complexity. Which MDM style fits best?
3. Why is it incorrect to think of MDM as simply 'a database or a system'?
4. Select ALL statements that are TRUE about what MDM is or isn't.
Select all the correct answers.
5. Select ALL correct statements comparing MDM implementation styles.
Select all the correct answers.
MDM in practice: the retailer case
A major European fashion retailer had accumulated 47 million customer records from 8 different systems over 10 years: two e-commerce platforms, a loyalty program, a wholesale portal, four regional POS systems, and a legacy catalog database.
Their problems were severe:
- The same customer appeared an average of 2.7 times across systems
- Different systems had different spellings of the same customer name
- Loyalty points couldn't be correctly allocated
- Marketing was spending 25% of its budget on duplicated contacts
- Customer lifetime valueCustomer lifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → calculations were systematically understated because purchases were split across duplicate records
Their MDM program (18 months, Reltio platform) consolidated 47M records into 17.5M Golden Records. The immediate outcomes:
- Marketing waste reduced by €4.2M annually
- CLV modeling became accurate for the first time (actual LTV was 31% higher than previously measured, because purchases were now correctly consolidated)
- Customer service resolution time improved as agents could see the complete customer history for the first time
Tool selection: the short version
The major MDM platforms: Informatica MDM (enterprise, complex implementations), Reltio (cloud-native, strong for customer data), Stibo Systems (strong for product data), SAP MDG (best if you're an SAP shop), Talend (open-source heritage, now part of Qlik).
Selection criteria: What is your primary use case (customer vs. product vs. supplier MDM)? What is your existing technology ecosystem? What is your integration architecture? How much organizational change can you manage in Year 1?
Don't select an MDM tool before answering these questions. The tool is the last decision, not the first.
What to do, from this lesson
These actions are compiled in the role's Playbook.
- Launch MDM on your highest-pain domain, choosing tools last
Related articles
Recent articles from the blog that build on this lesson.
- DataOne HCP, six records: why identity resolution is pharma's most expensive data problemA single cardiologist can exist as six different entities across a pharma company's CRM, claims data, and prescriber analytics systems, and none of them match. Until identity resolution works in practice, every downstream decision, from sampling allocations to pharmacovigilance reporting, is built on a fractured foundation.
- DataHow Salesforce learned to make master data stickSalesforce spent years selling data quality to its customers while quietly struggling with fragmented customer and product records across its own acquisitions. The way the company addressed that internal contradiction holds practical lessons for any CDO trying to move MDM from a slide deck into operating reality.
- DataWhen your AI strategy outpaces your data infrastructure: what CDOs must fix firstMany organizations are deploying AI models on top of data foundations that were never designed to support them. The performance gap this creates is not a technical footnote, it shapes whether enterprise AI delivers any measurable return at all.