Master data governance guide: step by step
A practical guide to set up master data governance: domains, roles, policies, approval workflow, metrics and a 30, 60 and 90 day plan.
A master data governance guide is the roadmap for deciding who is in charge of each master data domain, under which rules, and how those rules are enforced. The steps, in one sentence: define scope and domains, appoint roles and a decision forum, write the policies for each domain, put creation and change requests through an approval workflow, measure quality on a fixed routine, and sustain it all with a committee, periodic reviews and training.
The scene is familiar to almost every master data team. Someone urgently requests a new material, the analyst creates it, the purchase goes through. A month later the same item shows up with a different description, engineering says it never approved that standard, procurement says it only placed the request, and the rules spreadsheet that "someone built years ago" is nowhere to be found. Tools and goodwill are not what is missing. What is missing is a written agreement on who decides, what is mandatory and what happens to exceptions.
This guide is for whoever has to build that agreement inside the company. If your question is still conceptual, start with what master data is, the page on what MDM (master data management) is, or the difference between data management and data governance. Here the focus is the governance step by step (roles, forum, policies, workflow and routine); implementing the MDM program is covered in the MDM post.
What is master data governance in practice?
The DAMA-DMBOK, the data management body of knowledge maintained by DAMA International, defines data governance as the exercise of authority and control (planning, monitoring and enforcement) over the management of data assets. The same text notes that every organization already makes decisions about data, whether or not it has a formal governance function; the difference is doing it intentionally (DAMA-DMBOK, 2nd ed., Data Governance chapter, as reproduced by the DAMA Rocky Mountain Chapter).
Applied to master data (materials, services, products, suppliers, customers, assets), that authority becomes concrete: who approves a description standard, which field stops a record from moving forward, who releases a supplier with a pending document, who answers when duplicates go up. DAMA itself points out that the DMBOK does not prescribe a specific tool or methodology: each organization adapts the principles to its own environment (DAMA International, DAMA-DMBOK).
How do you set up master data governance step by step?
Step 1. Define the scope and domains of the first wave
List the domains that exist in the ERP and pick one or two to start. The most useful criterion is visible pain: where a master data error already costs money, delays invoices or blocks payments. Also define which companies, plants and systems are in scope. Governance that tries to cover everything in the first month usually turns into a document nobody uses. If you are torn between one domain or several at once, the article on single-domain or multidomain MDM helps you decide.
Step 2. Appoint the roles and create a decision forum
Each domain needs a data owner (who decides and is accountable) and at least one data steward (who looks after the rules day to day), plus the master data team that executes and IT that runs the system. The detailed split is in data owner vs data steward in master data, so it is not repeated here.
What is usually missing is the forum: the place where conflicts between departments are settled. Without it, a dispute such as "who defines the unit of measure, engineering or procurement?" ends up in email and stays without an owner. Write down:
- who takes part (owners of the domains in scope, a master data team representative, IT);
- who chairs the forum and records decisions;
- what only the forum can decide (policy changes, conflicts between domains, recurring exceptions);
- how a decision becomes a rule in the system and who confirms it was applied.
Step 3. Write the policies and rules for each domain
A master data policy does not need to be a treatise. For each domain it answers four questions: what is mandatory for the record to exist, what validation runs before saving, who approves what, and how an exception is requested, approved, recorded and closed.
For the data standard, use an explicit reference. ISO 8000-110:2021 deals with the exchange of characteristic master data and covers, among other points, conformance to a data specification and semantic encoding with access to data dictionaries. In practice, that means having an attribute dictionary per family or record type, like the one described in material attribute dictionary.
The table below is a template to start your policy matrix. The items are common examples; the right rule and approver depend on your company. Some items are Brazilian: the CNPJ is the Brazilian company tax ID, Receita Federal is Brazil's federal tax authority, and NCM is the Mercosur product classification code used on Brazilian invoices.
| Domain | Mandatory rule (example) | Validation before saving | Who approves | Allowed exception |
|---|---|---|---|---|
| Materials | Description following the family standard, unit of measure, NCM (in Brazil) and material group filled in | Duplicate search by similarity and attributes | Steward approves the standard; owner approves items outside the standard | Provisional record with a deadline to complete it, approved by the owner |
| Services | Structured service description, no free text, and a defined classification | Comparison with existing services | Services steward | One-off service with a recorded justification |
| Suppliers | Tax ID (CNPJ in Brazil), registration status checked with the tax authority (Receita Federal in Brazil) and qualification documents | Registration status and expired document checks | Procurement approves qualification; finance approves bank details | Emergency supplier with a pending document and a deadline to fix it |
| Customers | Tax data, address and commercial terms | Duplicate check by tax ID and address | Sales approves the record; credit approves the limit | Sale with an incomplete record for a set period |
| Any domain | Sensitive changes only when the requester is not the approver | Change history kept | Owner of the data group | None: segregation of duties has no exception |
Step 4. Put creation and change requests through an approval workflow
A rule that lives only in a document is not governance. The workflow is what makes the rule apply every day. For each domain, design:
- A request with minimum fields and a justification.
- Automatic validation of whatever a rule can check (mandatory field, format, duplicates, registration status).
- Steward review for anything outside the standard.
- Approval according to the policy matrix, with a deadline per step.
- Creation or change in the ERP, with the history recorded.
- Blocking, deactivation and reactivation following the same path, not a shortcut.
Also define what happens to urgent requests. If urgency has no path of its own, it becomes the main path.
Step 5. Pick a few metrics and set a review routine
A metric without a routine is a decorative dashboard. Choose three to five per domain, each with an owner and a target, and define when each one is reviewed. The metrics themselves (duplicates, completeness, cycle time, rejection at entry, incidents caused by master data) are explained in how to measure master data quality. What this guide adds is the routine. Below is a suggested template for the cadence:
| Forum | Suggested frequency | Who takes part | What comes out of the meeting |
|---|---|---|---|
| Master data team queue | Daily or weekly | Master data team and stewards | Pending items, stalled requests and expired exceptions |
| Quality review per domain | Monthly | Owner, steward and master data team | Metrics against target, root cause of deviations and actions with an owner |
| Governance committee | Quarterly | Owners, coordination and IT | Policy changes, conflicts between domains and the next wave |
Step 6. Sustain it: committee, periodic review and training
Governance breaks down when a manager changes, a new plant comes in or a rule gets old. To keep it standing:
- record owner and steward by job title as well as by name, with a named substitute;
- review policies at least once a year and whenever the system, regulation or company structure changes;
- train requesters, not only the people who create records: many errors start at the request;
- keep a log of committee decisions, with date and reason, so the same topic is not debated again;
- review open exceptions: an exception with no deadline becomes a parallel rule.
Checklist: is your master data governance in place?
- First-wave scope written down (domains, companies, plants and systems).
- Data owner and data steward appointed per domain, with substitutes.
- Decision forum with members, chair and decision rules defined.
- Policy per domain: mandatory fields, validations, approvers and exceptions.
- Attribute dictionary or description standard published and accessible.
- Workflow for creation, change, blocking and reactivation, with a deadline per step.
- Defined path for urgent requests, with a deadline to regularize.
- Segregation of duties for sensitive changes.
- Three to five metrics per domain, with owner, target and routine.
- Policy review calendar and training plan.
What are the most common mistakes when setting up governance?
- Starting with the tool. Software helps enforce rules, but it does not decide who owns the data.
- Scope too large. All domains in the first wave produce a generic policy and no real decisions.
- Roles without authority. Appointing an owner who cannot approve or hold their own department accountable.
- Policy without workflow. The rule exists on paper, but the system lets records through without it.
- Exceptions without deadlines. The provisional record is never completed.
- Metrics without meetings. The number exists, nobody looks at it and nobody acts.
- Cleaning the data without changing the entry point. Cleansing fixes the past; without governance, the data gets dirty again.
Hypothetical example: supplier governance at a manufacturer
Hypothetical example, created to illustrate the method. It does not describe a real customer.
A mid-sized manufacturer decides to start with the supplier domain, because payments have already been blocked by records with wrong bank details. The committee appoints the procurement manager as owner of supplier qualification and treasury as owner of the bank data group. The policy defines the mandatory documents per supplier type, a check of the supplier's tax registration status before saving, and the rule that whoever requests a bank account change cannot approve it. The workflow gets an "emergency supplier" step, with a deadline to regularize and owner approval. The monthly review tracks three numbers: active suppliers with expired documents, qualification lead time and bank changes made outside the workflow. Only after the routine has run for a few cycles does the committee open the second wave, with materials.
Where to start in 30, 60 and 90 days?
The split below is a suggested order, not a promise of timing or results. The real timeline depends on the size of the data, the number of departments involved and the availability of the owners.
- First 30 days: decide. Take the MDM Readiness Test to see which dimension of your master data is weakest, choose the first-wave domain, appoint owner and steward, and schedule the first forum meeting.
- By day 60: write and design. Write the policy for the chosen domain (mandatory fields, validations, approvers, exceptions), publish the attribute dictionary and design the creation and change workflow, including the urgent path.
- By day 90: measure and adjust. Take a baseline of the metrics, hold the first monthly review, fix the rules that did not fit real cases and decide in the committee whether the second wave starts or the first one needs another cycle.
If your company is close to an SAP S/4HANA project, cross-check this plan with the master data governance checklist for the S/4HANA kickoff.
Frequently asked questions
What is the difference between data governance and master data governance?
Data governance covers all of a company's data, including transactional, analytical and personal data. Master data governance is the subset that deals with records shared across departments and systems, such as materials, suppliers and customers. It uses the same principles, with rules per domain and per attribute.
Who should lead master data governance?
Coordination usually sits with a cross-functional area, such as the master data team, a data office or a process area. Decisions about each domain stay with the business owners. IT takes part as custodian of the system, without owning the data.
Do you need software to have master data governance?
Not to get started. Roles, policies and routines can exist before any tool, and the DAMA-DMBOK itself does not prescribe technology. Software starts to matter when volume grows and rules need to be applied in the workflow, with validation, approval and history, instead of relying on manual checks.
How long does it take to implement master data governance?
It depends on the number of domains, the current quality of the data and the availability of the departments. That is why the safest approach is to work in waves: one domain at a time, with the routine running before opening the next. Governance does not end; it becomes part of operations.
Do I need to clean the data before setting up governance?
The two go together. Without governance, cleansed data gets dirty again; without cleansing, the metrics start out very poor and discourage the committee. A common path is to define the domain policy, clean what is critical and switch on the entry workflow.
What is the difference between a master data policy and a procedure?
A policy says what applies and who decides: what is mandatory, who approves and when an exception is allowed. A procedure says how to carry it out: on which screen, in which order and with what evidence. The policy changes rarely and is approved by the committee; the procedure changes whenever the system or the process changes.
How 4MDG can help
Setting up governance usually runs into two questions: how big the problem really is, and who designs the model. 4MDG offers a free diagnostic, subject to an assessment of applicability and availability, on a real slice of your materials, suppliers and customers data, showing with numbers how much is duplicated, incomplete or irregular, with nothing installed in your environment. When the data keeps getting dirty because there is no policy, no defined roles and no tracked metrics, MDM Engineering designs the operating model: Master Data Management policy and strategy, data owner and data steward roles in a RACI matrix, a metrics map and the master data team model. To make the rules apply in the workflow, the platform has approval steps with an option for segregation of duties, deadlines per step, quality rules that prevent a record from being published and a versioned history, detailed in capabilities and architecture. If you want to discuss your case, get in touch.
Sources
The steps, templates and timelines in this guide are practice recommendations, not legal requirements.
- DAMA International. DAMA Data Management Body of Knowledge (DMBOK). https://dama.org/learning-resources/dama-data-management-body-of-knowledge-dmbok/. Accessed on November 21, 2026.
- DAMA Rocky Mountain Chapter. DMBoK Figure 14 Context Diagram: Data Governance and Stewardship. https://dama-rmc.wildapricot.org/news/13242927. Accessed on November 21, 2026.
- ISO. ISO 8000-110:2021, Data quality, Part 110: Master data: Exchange of characteristic data: Syntax, semantic encoding, and conformance to data specification. https://www.iso.org/standard/78501.html. Accessed on November 21, 2026.