What is Master Data Management (MDM)? Definition, Types, and Benefits

Updated October 2026

Master data management (MDM) is the practice of keeping one trusted, governed version of a company's core entities. Those entities are usually customers, products, suppliers, locations, and assets. MDM pulls records from many systems, matches duplicates, settles conflicts, and shares a golden record that every team and application can use. Programs differ mainly in how far the hub reaches into source systems.

What is master data management?

Master data management is a discipline and a set of tools for creating one trusted record for each customer, product, supplier, or other core entity. Informatica describes it as creating a single master record for each person, place, or thing. That record is drawn from internal and external sources, deduplicated, reconciled, and enriched.

IBM calls MDM a comprehensive approach to managing critical data across the enterprise. The software matters, but most of the work is agreement. Teams must decide what counts as a customer, who owns each field, and which system wins when two disagree. SAP stresses that MDM is an ongoing effort, not a one-time project.

What counts as master data?

Master data is the slow-changing core data that describes the entities a business deals with. IBM separates it from two other kinds of data. Transactional data records events such as sales, invoices, and claims. Reference data is the code sets used to classify other data, such as country and currency codes.

The split matters because every transaction points at master data. If one customer exists three times, every revenue report, churn model, and AI answer built on orders inherits that error. SAP notes that master data is a small share of business data but among the most complex and valuable. For how metadata and catalogs relate, see data catalog vs data governance vs data lineage.

How does master data management work?

MDM works as a loop of five steps. Records come in from source systems, get standardized, get matched into clusters, get merged into golden records, and go back out to the systems and people who use them. Stewards handle the exceptions that rules cannot settle.

1. Ingest. Pull customer, product, or supplier records from CRM, ERP, ecommerce, billing, and external data providers.

2. Standardize. Clean names, addresses, phone numbers, and codes into one format so records can be compared.

3. Match. Use rules, probabilistic scoring, or machine learning to find records that describe the same real entity.

4. Merge. Apply survivorship rules that pick the most trusted value for each attribute, producing the golden record.

5. Govern and share. Route low-confidence matches to stewards, keep an audit trail, and publish golden records to downstream systems.

Matching is the hardest step to get right, and it has its own discipline called entity resolution. The guide to entity resolution techniques covers blocking, scoring, and clustering. Survivorship rules, stable IDs, and steward queues are covered in identity resolution and entity mastering.

What are the four MDM implementation styles?

The four common MDM implementation styles are registry, consolidation, coexistence, and centralized. They differ in where the golden record lives and whether the hub changes source systems. Reltio's glossary describes each style as building on the one before it, with cost and accuracy rising together.

Registry style

The registry style cleanses and matches records from many sources and keeps a golden record in the hub as a read-only view. Source systems stay unchanged. It is quick and cheap to set up. The trade is that no human reviews the result, so Reltio warns the registry golden record is not 100% accurate.

Consolidation style

Consolidation adds a stewardship layer on top of registry. Algorithms still do the first pass, and data stewards review questionable matches and fix them. The result is accurate enough for reporting and analytics. Source systems still keep their own copies, so operational apps can drift from the golden record over time.

Coexistence style

Coexistence pushes golden records back into source systems, so master data lives in both the hub and the sources. People keep working in their usual tools and see cleaned data there. The cost is complexity. Each source needs its own quality checks, and the team must settle conflicts when the hub and a source change the same field.

Centralized style

In the centralized style, the hub is the only place where master data is created and changed. Stewards approve new records in the hub, and other systems subscribe to them. It delivers the highest accuracy and control. It also costs the most and changes daily workflows, because nobody creates a customer in the CRM directly anymore.

How to pick a style

Pick the lightest style that solves the problem in front of you. Registry or consolidation suits analytics, reporting, and a first trusted view. Coexistence suits teams whose operational apps must show clean data. Centralized suits regulated domains where every change needs approval. Many programs start light and move up one style at a time.

What are the main MDM domains?

An MDM domain is a category of master data managed with its own model, rules, and owners. Informatica lists customer, product, supplier, reference data, location, asset, and employee as common domains, plus industry entities such as patients, providers, and contracts.

Single domain vs multidomain MDM

Single-domain MDM manages one category, most often customer or product. Multidomain MDM manages several in one platform with one governance model, so teams can govern links such as which supplier provides which product. SAP Master Data Governance, for example, keeps one governed model across business entities while preserving the relationships between them.

When those relationships become the main question, a graph model often helps. The guide to how knowledge graphs transform master data management covers hierarchies, typed relationships, and graph matching in depth.

What is a golden record?

A golden record is the single best version of an entity, assembled from matching source records. Informatica calls it the best version of the truth. Survivorship rules pick each attribute value by source trust, freshness, or completeness. The golden record sits above source records and does not delete them.

That lineage matters. A steward who sees a wrong address must be able to trace which system supplied it and why it won. Good MDM tools keep an audit trail for every merge and let stewards split a bad merge without losing history.

How is MDM different from entity resolution, data governance, and a CDP?

MDM overlaps with three nearby disciplines, and each does a narrower job. Entity resolution finds which records match. Data governance sets the rules for all data. A customer data platform unifies customer profiles for marketing. MDM combines matching, governance rules, and stewardship to produce trusted records for many domains.

SAP describes governance and MDM as complementary, since governance rules attach to master data. Salesforce describes a customer data platform as software that builds a unified customer profile for segmentation and personalization. Many companies run both MDM and a customer data platform, with MDM supplying trusted identity. For governance tooling, see the data governance, lineage, and catalog tools comparison.

Which MDM vendors matter in 2026?

The MDM market consolidated in the last year, and AI agent access became a standard feature. Salesforce completed its Informatica acquisition on November 18, 2025. SAP closed its Reltio purchase on May 7, 2026 for 1,259 million dollars, according to its Q2 2026 statement.

Informatica MDM

Informatica, now part of Salesforce, says it was named a Leader in the 2026 Gartner Magic Quadrant for MDM for the seventh time. Its matching and stewardship tools are mature and widely deployed. Salesforce plans a single data pipeline with MDM on Data 360, so buyers outside Salesforce should watch future packaging.

Reltio

Reltio announced Leader placement and says it was positioned furthest for Completeness of Vision. SAP says Reltio will become a core capability within SAP Business Data Cloud and credits its MCP support for agent workflows. SAP also says Reltio stays available standalone for the foreseeable future.

SAP Master Data Governance

SAP Master Data Governance is the governance layer for master data inside SAP Business Data Cloud. It offers matching and merging, quality rules, change request workflows, and a full audit trail. Its built-in integration with SAP's data model is its main strength, and also why it fits best where SAP already runs the core business.

Profisee

Profisee announced Leader placement in April 2026 and the general availability of a Fabric-native MDM experience, plus an MCP server for agents. Its pricing is volume based with unlimited domains, and it runs as SaaS on Azure or as containers on other clouds. Prices require a quote.

Semarchy

Semarchy also announced Leader placement in 2026 for the Semarchy Data Platform. It runs as SaaS, self-hosted, or as a Snowflake Native App, with Git and CI/CD support. A 30-day free trial is available, but paid pricing is not listed.

Stibo Systems STEP

Stibo Systems announced Leader placement and calls itself the largest independent MDM provider. Its MCP server, launched June 2026, lets agents query governed master data with its relationships and hierarchies. The release does not say whether agents can write data back.

Tamr

Tamr focuses on AI-driven matching and charges per golden record with unlimited users. It also lists an Enterprise Knowledge Graph feature and MCP connectivity. Tamr publishes no prices and offers a guided proof of concept rather than a free trial.

Why does master data matter for AI?

Master data matters for AI because agents and models can only be as consistent as the entities they reason about. An agent that cannot tell three customer records apart will count one customer three times. Stable IDs, agreed definitions, and clean hierarchies are what let an agent answer revenue or risk questions correctly.

Vendors now expose master data to agents directly. Reltio, Stibo Systems, Profisee, and Tamr each document Model Context Protocol support, as cited above. Master data still needs the business meaning around it, which is where ontologies and semantic layers come in. See how ontology powers AI analytics for that layer.

Who should own your master data model?

Your company should own its master data model, because it encodes how the business defines customers, products, and partners. That knowledge outlasts any vendor contract, and the recent acquisitions show how fast ownership of MDM tools can change. Keep match rules, entity IDs, and definitions documented in open formats you can move.

A partner can help build that foundation on the stack you already run. Galaxy is an applied lab that does this work with internal tooling for semantic models, ontologies, and entity resolution, and it builds in the open with Filament by Galaxy.

How do you start an MDM program?

Start with one domain and one painful business question. Customer and product are the most common first domains. A narrow start proves the matching rules, the owners, and the steward workflow before money goes into more sources.

1. Pick one domain and name a business owner for it.

2. Write down the questions the business cannot answer today because of duplicate or conflicting records.

3. Agree on definitions, such as what counts as an active customer.

4. Choose the lightest implementation style that answers those questions.

5. Profile the source systems, then build and test match and survivorship rules.

6. Set up steward queues and measure match quality before adding the next domain.

Frequently asked questions

What is master data management in simple terms?

Master data management is the practice of keeping one trusted, governed version of the core things a business deals with, such as customers, products, suppliers, and locations. It pulls records from many systems, removes duplicates, fixes conflicts, and shares the result. Everyone then works from the same customer or product instead of several competing copies.

What is the difference between master data and transactional data?

Master data describes the entities a business deals with, such as a customer or a product, and it changes slowly. Transactional data records events, such as a sale, an invoice, or a claim, and it grows every day. Each transaction points at master data, so errors in master data spread into every report built on transactions.

What are the four MDM implementation styles?

The four common styles are registry, consolidation, coexistence, and centralized. Registry matches records and leaves sources alone. Consolidation adds human stewards who review the golden record. Coexistence pushes the cleaned record back into source systems. Centralized makes the MDM hub the only place where master data is created and changed, which gives the most control at the highest cost.

Which MDM implementation style should I start with?

Many teams start with registry or consolidation because they leave source systems unchanged and deliver a trusted view for analytics quickly. Move to coexistence when operational systems need the cleaned data back. Choose centralized only when the business is ready to create and approve master data in one place, since it changes how people work every day.

What is multidomain MDM?

Multidomain MDM manages several master data domains, such as customers, products, suppliers, and locations, in one platform with one governance model. It lets teams govern the links between domains, such as which supplier provides which product. Most programs still start with one domain and add others once matching, stewardship, and ownership work for the first.

What is a golden record in MDM?

A golden record is the single best version of an entity, built by matching records from several systems and picking the most trusted value for each attribute. Informatica calls it the best version of the truth. It sits above source records rather than replacing them, so teams can always trace each value back to the system it came from.

How is MDM different from data governance?

Data governance sets the policies, owners, quality standards, and access rules for all data. MDM applies those rules to a small set of critical entities and produces trusted records from them. SAP describes the two as complementary. Governance without MDM leaves duplicate customers in place, and MDM without governance has no agreed rules to enforce.

How is MDM different from a customer data platform?

A customer data platform unifies customer profiles mainly for marketing, such as segmentation and personalization. MDM governs many domains, including products and suppliers, and focuses on accuracy, stewardship, and audit trails. Many companies run both. MDM supplies trusted customer identity and core attributes, and the customer data platform adds behavior data and activates profiles in campaigns.

Which MDM vendors lead in 2026?

Informatica, now owned by Salesforce, Reltio, now owned by SAP, Profisee, Semarchy, and Stibo Systems each announced Leader placement in the 2026 Gartner Magic Quadrant for MDM. SAP Master Data Governance is SAP's own option for SAP landscapes, and Tamr focuses on AI-driven matching. Each vendor has a different deployment model and pricing approach.

Does MDM matter for AI agents?

Yes. Agents need stable entity IDs and agreed definitions to answer questions about customers, products, or suppliers correctly. Without them, an agent may count one customer three times. Reltio, Stibo Systems, Profisee, and Tamr all document Model Context Protocol support, so agents can query governed master data instead of raw tables.