Master data management (MDM) is the discipline in which business and IT work together to ensure the uniformity, accuracy, stewardship, semantic consistency and accountability of the master data shared across the company. It is enabled by technology but also involves people and processes, and provides the procedures for collecting, aggregating, matching, consolidating, quality-assuring, persisting and distributing master data (Wikipedia). Master data are the products, accounts and parties with which the business completes its transactions: customers, suppliers, products, employees.
The pages that dominate this topic mostly describe tools. The problem, however, arises before the software: in a company with several systems, the same customer exists in multiple copies, and no one has established which one counts. Who decides what “the customer” is, who can change it and which system prevails in a conflict is a matter of governance, which technology is then called upon to enforce.
Master data and transactional data: what MDM governs
Master data describes the entities around which the business operates and changes rarely; transactional data records what happens to those entities and is generated continuously. Master data management deals with the former, because an error in a master record is repeated in every transaction that uses it. The table shows the most common domains and the governance question each one raises.
| Domain | Sample master record | Question that governance must answer |
|---|---|---|
| Customers | Company name, VAT number, addresses, contacts | When are two records the same customer? |
| Products | Code, description, hierarchy, unit of measure | Who approves a new code, and with what minimum attributes? |
| Suppliers | Core details, terms, bank details | Who can change bank details, and under what controls? |
| Employees | Registry and organisational data | Which system is the source, and who can correct it? |
| The right-hand column is what separates a governance project from an IT project: these are questions no software can answer on the company’s behalf. | ||
Why master data management is needed
Master data management addresses, first and foremost, a problem of inconsistency: multiple separate copies of the same data produce different values, and decisions taken on different data cannot be compared. Suppose a bank sends mortgage advertising to someone who has already taken one out, because the two records have not been linked. The second problem is redundancy: the same entity appears across different product lines and creates confusion.
The third is mergers and acquisitions, where reconciling master data systems is complex: the cited entry refers to organisations with dozens or even a hundred poorly integrated master data repositories. The objectives of master data management are a single version of the truth, operational efficiency, decision support and regulatory compliance. Where master data affects risk, for example when an error concerns a supplier or a regulated customer, the issue also touches on risk management.
The golden record and the master data lifecycle
The golden record is the consolidated, authoritative version of an entity, built from the copies scattered across different systems. It is produced through a cycle that Wikipedia describes in stages: source identification, collection, transformation and normalisation, error detection and correction, consolidation, storage and distribution, with hierarchy management and governance overseeing the whole. The tools include de-duplication, standardisation and rules that prevent incorrect data from entering.
Making it operational takes three decisions:
Decision 1
Identity
Which attributes determine that two records are the same party.
VAT number, tax code, combinations of name and address.
Decision 2
Survivorship
When two systems hold different values for the same field, which one prevails.
And above all: why.
Decision 3
Refresh
Where the data is changed, and how the change reaches the other systems.
One point of entry, many points of access.
- Identity rule. Which attributes determine that two records refer to the same entity: VAT number, tax code, combinations of name and address.
- Survivorship rule. When two systems hold different values for the same field, which one prevails and why.
- Update rule. Where the data is changed and how the change reaches the other systems.
Without these rules, the golden record remains a good intention, and master data management a software project. With them, every conflict has an answer that is decided in advance and documented.
Master data management implementation models
Let us now draw an important distinction between the reference source and the distribution method. In the first case a single application, database or spreadsheet is the authoritative source (source of record): it is simple, but may not suit complex distributions, and it can be federated by attribute group or geographical area. For distribution, by contrast, three methods should be listed:
- consolidation, in which data flows into a central repository that replicates it;
- federation, which offers a single virtual view of the data while leaving it in the sources;
- propagation: copying data from one system to another, usually through point-to-point interfaces.
Which master data management model to choose depends on how many systems there are, how autonomous they are and who needs to be able to change the data. The choice concerns the organisation before the architecture: if two functions refuse to give up control of a master record, no technical model will solve it.
Who governs master data: data owners and data stewards
In master data management, Wikipedia describes two roles. The data owner sets the requirements for data definition, quality and security and ensures compliance with governance procedures; they also fund improvement work when deviations occur. The data steward runs MDM on the owner’s behalf and usually acts as their adviser. Each role may be assigned to several people, each responsible for a subset of the data: one owner for customers, another for employees.
Master data management is a discipline defined by the policies and procedures of the data governance organisation: without data governance assigning roles and rules, MDM lacks authority. This is why the subject is linked to the role of the data owner, that of the data steward and data quality management, which we cover separately, as well as to tracing where data comes from, the subject of data lineage.
Master data quality and exchange: the ISO 8000-110 standard
There is a specific standard for exchanging master data between organisations and systems: ISO 8000-110:2021, second edition, published on 15 November 2021. It sets requirements that can be verified with IT tools for messages exchanging characteristic master data: formal syntax, semantic encoding, conformance to data specifications and access to the dictionaries needed to decode them. It does not cover the internal management of master data or its accuracy and completeness, which are addressed by other parts of the series, and the standard states that its requirements are necessary but not sufficient to guarantee the quality of the exchange (ISO).
For anyone practising master data management the message is practical: a clean internal master record is not enough if the data exchanged with customers and suppliers lacks a shared format and meaning. Trust in data among partners and stakeholders also depends on this.
The Bliss angle: the question “who is the customer?”
The first meeting of a master data management project should ask a deceptively simple question: what, for us, is a customer?
Sales
The contract
Administration
The VAT number
Marketing
Who opens the emails
These are legitimate, differing answers, and software can accommodate them all only once someone has decided which definition serves as the reference and which remain attributes.
This is why we see master data management as an exercise in shared decision-making. It moves forward when there is a named owner for each domain and a written criterion for resolving conflicts. It stalls when the definition remains ambiguous and each function keeps its own master records “just in case”.
In that case you buy a platform and end up with one more copy.
The Bliss Agency perspective
Domande frequenti
What is master data management?
Master data management is the discipline in which business and IT work together to ensure the uniformity, accuracy, stewardship, semantic consistency and accountability of the master data shared across the company. It is enabled by technology but also involves people and processes, and sets out the procedures for collecting, consolidating, controlling and distributing master data (Wikipedia).
What is master data?
They are the products, accounts and parties with which the business completes its transactions: for example customers, suppliers, products and employees, together with their hierarchies (Wikipedia). They change rarely and are used by many systems and functions. Unlike transactional data, which records events, they describe the entities around which the company operates.
What is MDM for?
It serves to eliminate inconsistent versions of the same data, reduce redundancy and make decisions taken by different functions comparable. It also helps in mergers and acquisitions, where many master data repositories must be reconciled. The stated objectives are a single version of the truth, operational efficiency, decision support and regulatory compliance (Wikipedia).
What is a golden record?
The golden record is the consolidated, authoritative version of an entity, such as a customer, derived from the copies held in different systems. Building it requires identity rules, to recognise duplicates, and survivorship rules, to select the correct value when systems disagree. Without these rules it remains a stated objective.
What is the difference between MDM and data governance?
Data governance defines roles, rules and responsibilities for data across the entire company; master data management applies those rules to a specific set of data, the shared master records. Wikipedia describes it as a discipline defined by the policies and procedures of a data governance organisation: the latter gives authority to the former.
What is the difference between MDM and data quality?
Data quality concerns how fit any data is for its intended use. MDM works on master data and includes quality checks and error correction, but its purpose is also to create a single version and manage its distribution. Good quality is not enough if several competing records exist for the same entity.
Who is responsible for master data within the company?
Typically two roles: the data owner, who sets requirements for definition, quality and security and ensures compliance with governance procedures, and the data steward, who runs MDM on the owner’s behalf. Each role can be shared among several people, each responsible for a subset of data, for example customers or employees.
What are the implementation models for MDM?
A distinction is drawn between a single reference source, which can be federated by attribute or geographical area, and the various distribution methods: consolidation into a central repository, federation with a single virtual view, and propagation between systems via point-to-point interfaces. The choice depends on the number of systems, how autonomous they are and who needs to change the data.
Is MDM a piece of software?
No, software is only one part. MDM is a discipline that encompasses people, processes and technology: it defines who decides on each master record, under which rules and with which controls. A tool can automate de-duplication, consolidation and distribution, but without shared roles and rules it will not produce a reliable version of the data.

