A supplier created three times, with three different spellings, generates three invoices, three manual bank reconciliations, three lines in the supplier reporting, and an accounting error discovered on the 3rd of the month by a management controller who spends the day trying to understand why the totals don’t match. This scenario isn’t theoretical. It’s the daily reality for most mid-sized companies that deploy an ERP without first structuring their master data. Master Data Management (MDM) isn’t a technical option reserved for FTSE 100 companies. It’s the discipline that determines whether your ERP will last ten years or collapse under its own inconsistencies three years after go-live.
The Essential in 30 Seconds
- MDM isn’t a tool, it’s a discipline: you can do it with Excel and strict rules, you can’t skip it without paying the debt later.
- Retrospective deduplication costs 5 to 10 times more than prevention at creation, according to project feedback.
- Integrated MDM solutions (SAP MDG, Oracle DG, Microsoft Dataverse) have caught up: viable for most single-vendor mid-caps.
- Dedicated MDM (Informatica, Stibo, Semarchy, Reltio, Profisee) remains the reference when sources are heterogeneous or volume is critical.
- Governance trumps tools: well-governed Excel beats Informatica MDM without a data steward.
- The regulatory calendar (GDPR, CSRD, e-invoicing 2027) de facto imposes minimal MDM on customer and supplier repositories.
Master Data Management: What Are We Really Talking About?
An Operational Definition, Not a Gartner Definition
Master Data Management refers to the set of processes, roles, rules, and tools that ensure a single business entity (a customer, a product, a supplier, an employee) has a unique, consistent, and authoritative representation across all enterprise information systems. Simply put: when the sales rep opens the “Smith Ltd” file in the CRM, the accountant sees exactly the same file with the same company name, the same registration number, and the same customer code as the management controller in the ERP and the logistics manager in the WMS.
It’s not a one-off IT project. It’s an ongoing organizational capability.
The Four Master Data Domains
Four families concentrate 90% of ERP project problems:
- Customers (accounts, prospects, contacts): duplicates, inconsistent addresses, poorly modeled group hierarchies, B2B vs B2C deduplication.
- Products or materials (references, nomenclatures, units of measure): divergent coding between marketing, purchasing, and production, missing mandatory attributes.
- Suppliers (purchasing parties, service providers): duplicate bank details, erroneous registration numbers, obsolete banking data that opens the door to wire fraud.
- Employees and structures (collaborators, cost centers, org charts): critical for payroll, analytical reporting, and access governance.
Banks and insurers add a fifth domain (accounts, contracts), industry adds a sixth (equipment, technical locations). But for an industrial or service mid-cap, these four concentrate the main effort.
Master Data, Transactional, Reference: The Distinction That Matters
Many projects fail to establish this distinction. Transactional data are facts (an order, an invoice, an accounting entry): they’re written once and stack up. Reference data are stable nomenclatures (ISO country codes, currencies, general chart of accounts): they’re shared, rarely modified, often external. Master data are business entities that evolve over time and participate in transactions: they must be governed, versioned, and subject to a golden record.
Confusing these three families means ending up with a chart of accounts in the same repository as customers. And vice versa.
Why ERP Projects Without MDM Fail at Three Years
The Typical Scenario: Duplicates, Inconsistent Nomenclatures, Rejected Invoices
An industrial group of 1,200 employees starts their ERP in January. Migration of third parties from three legacy sources: Salesforce CRM, the old SAP system, and an Excel file maintained by purchasing. Nobody wrote matching rules. Result: 43,000 customers in the database, but in reality 28,000 distinct entities. Eleven months later, the new CFO discovers that the top 50 customers in reporting doesn’t match the top 50 in the customer service database, because “Johnson Controls” exists under six different customer codes.
Multiply by four domains and by as many acquisitions or functional redesigns, and you get an ERP that technically works but no longer delivers on business promises. BI teams build reports by circumventing repositories. Finance services maintain their own Excel spreadsheets. The ERP project becomes a fixed cost that produces less and less value.
The Real Costs of Data Debt
Experian estimates in their Data Governance Report 2024 that 78% of surveyed organizations have an executive sponsor for data governance, but only 69% have a dedicated team—a gap that betrays often symbolic governance. In projects, data debt is paid in hidden time: between 20 and 30% of BI and management control teams’ time is absorbed by reconciliation and cleanup, according to recurring feedback from post-go-live audits. The IBM study cited since 2016 quantified approximately $3.1 trillion annually the cumulative cost of poor data quality for the US economy, an order of magnitude still frequently referenced in literature, to be handled with the precautions of an old reference figure (source compiled by SAP Community).
The Ratchet Effect: The Longer You Wait, The More Expensive It Gets
Correcting a customer duplicate at creation costs a data steward a few seconds. Correcting it three years later means: reopening past orders, redoing invoices, redoing bank reconciliations, reclassifying accounting entries, possibly re-issuing credits, correcting consolidated dashboards, and documenting all this for auditors. Field feedback converges on a 5 to 10 ratio between late correction cost and prevention cost. This is the economic argument that justifies investing in MDM before go-live.
Integrated ERP MDM vs. Dedicated MDM: The Dividing Question
MDM Integrated with ERP Platforms
SAP Master Data Governance (MDG) is the native option for SAP S/4HANA environments. It covers the four major domains, offers pre-configured approval workflows, and integrates natively with SAP Business Data Cloud. Its main historical flaw: strong functional prescription that suits matrix organizations poorly.
Oracle Data Governance and Oracle Customer Hub cover MDM needs for Oracle Fusion Cloud ERP environments. The strength: native integration with Oracle business processes. The limitation: coverage of non-Oracle sources remains perfectible.
Microsoft Dataverse (formerly Common Data Service) plays the role of shared data platform between Dynamics 365, Power Platform, and Azure. It’s not strictly speaking an MDM in the Gartner sense, but it handles a large part of MDM use cases for organizations standardizing on the Microsoft stack.
Dedicated, Best-of-Breed MDM
The 2026 Gartner Magic Quadrant for Master Data Management Solutions, published April 6, 2026, evaluates 20 vendors and positions notably Salesforce (Informatica), Profisee, Semarchy, and Reltio in the Leaders quadrant (Profisee announcement, Semarchy announcement). Stibo Systems is also evaluated and remains a historical reference on product MDM, notably in distribution and industry. Ataccama progresses on the augmented data quality segment. SAP is evaluated but its positioning in the 2026 MQ primarily reflects the historical MDG portfolio. The Reltio acquisition, announced in late March 2026 and analyzed in detail here, will redistribute the cards on the next cycles.
Concrete Selection Criteria
Four questions help decide in 80% of cases:
- Source heterogeneity: if your master data comes from more than three non-ERP source systems (CRM, PIM, vertical business tools, M&A applications), dedicated MDM is justified. If everything comes from a single ERP with some extensions, integrated MDM suffices.
- Volume and frequency: beyond 500,000 active customer entities or 100,000 product references, dedicated MDM performance becomes an argument.
- Governance maturity: powerful MDM on an ungoverned organization remains useless. If you haven’t identified your data owners or formalized your approval workflows, start with integrated MDM while maturing.
- Budget: cloud dedicated MDM typically starts between $150-400K annually for a mid-cap, licenses and initial implementation. Integrated MDM reuses existing ERP licenses, the marginal cost is mainly implementation and governance.
Methodology: Structure Your Master Data in Six Steps
Step 1: Source of Truth Mapping
By domain (customers, products, suppliers, employees), exhaustively identify each system currently hosting the data, its estimated quality level, the process feeding it, and the functional owner. Expected output: a source × domain × quality × owner matrix. Without this diagnosis, no matching rule makes sense.
Step 2: Golden Record Definition by Domain
For each domain, formalize what constitutes the reference: which attributes are mandatory, which sources have priority in case of conflict, which survivorship rules apply. The golden record is not an average. It’s the best value from sources, validated by a data steward, according to an explicit survivorship rule (e.g., “priority to CRM data for name, to ERP for registration number, to purchasing tool for banking coordinates”).
Step 3: Matching and Deduplication Rules
Define rules that identify when a record from source X and a record from source Y refer to the same entity. Example for UK B2B customers: strong matching on Companies House number, medium matching on company name + postal code + house number, weak matching on trading name. These rules produce three result categories: automatic merges, manual review suspensions (data steward decides), rejections. Test rules on a sample before any industrialization.
Step 4: Creation and Modification Workflows
Every creation or modification of a master record must go through a formalized workflow: initial entry, automatic enrichment (Companies House number verification via API, anti-fraud scoring, IBAN check), business approver validation, propagation to consuming systems. Without workflow, your upstream efforts are nullified by wild creations from the sales department at quarter-end.
Step 5: Quality KPIs
Managing what you don’t measure is wishful thinking. Four minimum indicators per domain:
- Active duplicate rate (target: < 0.5% on customers, 0% on active suppliers).
- Completeness rate on mandatory attributes (target: 100% on products, 98% on customers).
- Average data freshness (age of last update).
- Average creation time for a validated record (measures workflow efficiency).
Step 6: Continuous Governance Runbook
MDM is never finished. A runbook documents rules, escalation procedures, RACI matrix, monthly quality reports, quarterly matching rules reviews, annual cleanup exercises. Without a runbook, MDM erodes in 18 months.
Governance: Who Does What, and Especially Who Validates
The Data Steward Role: How to Create It Without Hiring
The data steward isn’t necessarily a new job. In a mid-cap, it’s often a role carried at 30-50% by a senior employee from a business department: a key account manager for the customer domain, a purchasing manager for suppliers, a product reference manager for the catalog. This role must be named, formalized in the job description, integrated into annual objectives, and recognized by management. Without explicit recognition, it disappears at the first operational overload.
MDM Committee: Frequency, Participants, Deliverables
A monthly MDM committee, one hour maximum, with data owners from the four domains, the CIO (or their data representative), and a general management representative (CFO most often). Standard agenda: quality indicators, reported incidents, rule arbitrations, new evolution requests. Deliverables: documented decisions, updated quarterly roadmap.
Quality SLA by Domain
Formalize governance service commitments to business departments. Examples: validated customer creation within 4 business hours, urgent modification within 2 hours, active customer duplicate rate maintained below 0.5%, product completeness at 100% on mandatory attributes. These SLAs must be published, monitored, and made public in the organization.
MDM and Regulations: GDPR, CSRD, E-Invoicing
GDPR: Minimization and Right to Erasure
The GDPR framework applied to ERP imposes two strong constraints on customer master data: minimization (only collect what’s necessary and documented by a purpose) and right to erasure (ability to delete or anonymize an individual on request, across all systems). Without MDM, the right to erasure becomes an operational nightmare: you have to find all records of the same person across all systems, without certainty of exhaustiveness. With MDM and a propagated unique identifier, it’s one query and one procedure.
CSRD: Supplier Data Becomes Critical
The CSRD and sustainability reporting impose for scope 3 emissions traceability by supplier, even by purchasing category. An incomplete or inconsistent supplier repository makes CSRD reporting impossible to audit. This is one of the levers justifying MDM investment now, even if the organization didn’t perceive the urgency before.
E-Invoicing 2027: Supplier Repository Quality as a Prerequisite
The generalization of e-invoicing in Europe transforms each supplier into a node in an invoicing network (Peppol, Chorus Pro, KSeF, ZUGFeRD by country). Note that the UK has implemented Making Tax Digital (MTD) as its equivalent framework, requiring similar data quality standards for VAT reporting and supplier validation. An erroneous VAT number, an obsolete intra-community VAT number, or a poorly formatted address generates automatic rejections. Supplier repository quality becomes an operational condition, not just comfort.
MDM Checklist for an Ongoing ERP Project
- The four master data domains are identified with their named data owner.
- Source systems mapping is documented and dated less than 6 months ago.
- Golden record rules by domain are formalized and validated in committee.
- Matching and deduplication rules have been tested on a representative sample.
- Creation and modification workflows are tooled and go through business validation.
- Registration numbers and VAT numbers are automatically verified at entry.
- Quality KPIs are published monthly and accessible to management.
- At least one data steward per domain is named and has the necessary time.
- A monthly MDM committee meets with quorum and produces documented decisions.
- GDPR minimization and erasure rules are implemented on customer and employee domains.
- CSRD reporting processes rely on the governed supplier repository.
- A continuous governance runbook is maintained and reviewed at least quarterly.
Twelve points to evaluate MDM maturity. Seven out of twelve validated, the ERP project has a chance. Below five, go-live is a gamble.
Going Further
MDM isn’t a topic you handle once and forget. It’s a capability built alongside ERP processes that reveals itself in use. To go deeper, read our complete CRM-ERP integration guide which details flow architectures between customer repositories and transactional systems, our analysis of SAP’s acquisition of Reltio which illuminates major vendors’ strategy on the data layer, and our ERP and Business Intelligence guide which concretely shows what MDM makes possible on the reporting side.
If you’re wondering where to start with your own master data, the answer fits in one sentence: name a data owner per domain in the next 30 days, and convene a first MDM committee in the next 60 days. The rest is work.
