All services Every engagement has a fixed, published price. Book a strategy call
The practical difference between data governance and data management, a test you can apply in a meeting, how to match your symptom to the discipline that fixes it, and which to fund first.
Nobody asks this question out of curiosity. It gets asked when a budget request has landed, or two reports disagree, or an AI project has stalled, and the person deciding cannot tell whether they are being sold a platform or a set of rules about who may decide things.
The usual answers do not help, because they are written as glossary entries. Governance is about policy and management is about execution, which is true and useless. It tells a finance director nothing about whether the correct next step is a tool, a hire, an operating model or a data cleanup.
This guide answers the question the way a leader needs it answered. Why it comes up, the one line distinction and a test you can apply in a meeting, what sits on each side, how to match the symptom you are experiencing to the discipline that fixes it, who owns what, which to fund first including the cases where the platform work genuinely comes first, and how the two are treated differently by regulators. Every fee named here is published on our own service pages.
The short answer. Data management moves, stores, models and serves data. Data governance decides who is allowed to decide things about it, including definitions, ownership, access, quality thresholds and retention. The quickest test is to ask what the answer to a question is: if the answer is a named person, you are in governance, and if the answer is a tool or a pipeline, you are in management. Most organisations have bought management and skipped governance, which is why they own several reporting platforms and still cannot agree what a customer is. Fix the discipline that matches your symptom rather than the one your supplier sells. Our governance work starts from £14,500 + VAT for a ten day assessment and from £45,000 + VAT to build the operating model.
The trigger matters, because each of these five situations has a different correct answer. Read down the first column and find the one that describes your week.
| What prompted the question | What is really being asked | The honest answer |
|---|---|---|
| Two reports gave different numbers to the board | Do we need a better platform or better discipline | Discipline. Nobody owns the definition, so both reports are correct against different rules |
| An AI project has been blocked on data concerns | Which team unblocks this | Governance decides whether the data may be used and who says so. Management supplies it once that is settled |
| An audit or a regulator asked a question nobody could answer | Is this a compliance job or a technology job | Governance, almost entirely. The answer required is ownership and evidence rather than a system |
| A migration or an ERP programme is slipping on data | Do we stop and govern first | No. Profile and fix the sources now, and stand up ownership alongside so the same mess does not arrive in the new system |
| A new data leader has asked for budget | Is this empire building | Ask what decisions the money closes. A credible request names decisions and owners rather than a technology stack |
Data management is the work of moving, storing, modelling, protecting and serving data so that people can use it. The DAMA body of knowledge sets out the recognised knowledge areas, and most of them are engineering and operational disciplines.
Data governance is the work of deciding who may decide. Which definition is the approved one, who owns a domain, who grants access, what quality threshold is acceptable, how long records are kept and who is accountable when any of that goes wrong. ISO/IEC 38505 treats governance of data as a governing body responsibility rather than a technical one, which is the right instinct.
Here is the test worth carrying into a meeting. Take any real question about your data and ask what form the answer takes. If the correct answer is a person, you are in governance. If the correct answer is a tool, a table or a pipeline, you are in management.
Run this on the five questions that annoy you most. The split is usually obvious within a minute, and the pattern normally shows a management estate that has grown without a governance layer above it.
| The activity | Which discipline | Who usually holds it |
|---|---|---|
| Agreeing the approved definition of a business term | Governance | A named business data owner |
| Building and running ingestion, storage and modelling | Management | Data or platform engineering |
| Deciding who may see or use a dataset | Governance | The data owner, advised on legal grounds |
| Implementing the access controls once decided | Management | Engineering and identity teams |
| Setting the acceptable quality threshold for a field | Governance | The data owner who lives with the consequence |
| Measuring quality and reporting the breaches | Management | Engineering, with results going to the owner |
| Deciding what is kept, for how long, and what is deleted | Governance | The owner, with data protection advice |
| Executing deletion and archiving reliably | Management | Engineering and infrastructure |
| Selecting a catalogue, warehouse or integration platform | Management, with governance requirements as input | Technology, informed by owners |
| Resolving a dispute between two teams about a number | Governance | The escalation route, ending with a named decider |
Two things are worth noticing. Almost every governance row ends in a person and almost every management row ends in a mechanism. And they interlock at every line, which is why treating them as competing budget requests produces bad decisions.
Leaders arrive with symptoms rather than vocabulary. This table is the practical core of the article. Find your symptom, then read the last column carefully, because that is where most money gets wasted.
| The symptom | Which discipline | What fixes it | What will not fix it |
|---|---|---|---|
| Two reports give the board different revenue figures | Governance | One named owner per definition, the definition written down, one approved source declared | Another dashboard, or replacing the reporting tool |
| A report takes three days to refresh | Management | Pipeline engineering, scheduling and monitoring | A governance policy or a steering committee |
| Nobody can say whether a dataset may be used to train a model | Governance | Classification, a written rule and an approval route with a named decider | Adding metadata to a catalogue nobody is accountable for |
| Customer records are duplicated across systems | Both, in this order | Governance names the surviving record and its owner, then management builds the matching and merge | Buying a master data platform before anyone owns the rule |
| An audit asked who has access to personal data and nobody could answer | Governance | An ownership register, access decision rights and evidence that both are applied | A one off permissions cleanup that nobody maintains |
| A migration keeps slipping because the source data is worse than expected | Management first | Profiling and remediation on the real fields, with owners appointed alongside | Pausing the programme to write a strategy |
| The AI pilot works on sample data and fails in production | Both | Quality measurement on the exact fields the model consumes, plus an owner for each of them | More model tuning |
| Storage costs climb and nobody can say what may be deleted | Governance | Retention decisions made by owners and written into a schedule | Moving everything to cheaper storage |
| Every reporting request goes through the same two people | Management, with a governance cause | Self service models and documented sources, once definitions are agreed and owned | Hiring a third analyst into the same bottleneck |
The pattern in the last column is consistent. Governance problems get treated with technology purchases, and management problems get treated with committees. Both substitutions are expensive and both are common.
Most confusion between the two disciplines is really a gap in this table. Read it and identify the row you cannot name a person for.
| Role | On the governance side | On the management side |
|---|---|---|
| Board | Confirms the data the organisation will be held to, and receives assurance that ownership exists | Approves platform investment on a business case |
| Chief data officer or executive sponsor | Owns the operating model, appoints data owners, resolves disputes | Sets platform direction and holds the budget |
| Business data owner | Approves definitions, decides access, accepts quality thresholds and retention | Specifies what the business needs built |
| Data steward | Keeps definitions and the register current, chases exceptions | Works with engineers so changes reach the models |
| Data or platform engineering | Implements the controls that governance has decided | Builds and runs ingestion, storage, modelling, quality measurement and access |
| Data protection officer | Advises on lawful basis, retention and individual rights, and checks the evidence | No operational role in the pipeline |
In organisations where this question causes arguments, the missing row is almost always the business data owner. Engineering has been asked to decide what a customer is, which is a decision engineering has no standing to make and no way to defend.
Governance consultancies rarely say this, so it is worth saying plainly. There are cases where management should be funded first, and pretending otherwise wastes a year.
Governance goes first when numbers are disputed, when an AI or analytics initiative is blocked on permission rather than plumbing, when an audit finding is open, when two teams both believe they own the same data, or when nobody can say what may be deleted. In each of those, more technology makes the problem harder to see.
If the honest answer is that you cannot tell which situation you are in, that is what a short assessment is for. Our Data Governance QuickScan runs for ten working days from £14,500 + VAT and it is designed to answer exactly this question with evidence rather than opinion.
These two disciplines are bought from different suppliers on different commercial models, which is part of why they get conflated. Management is usually bought as licences plus implementation days, so the cost is visible and the value is assumed. Governance is bought as people and decisions, so the cost is visible and the value is questioned.
| What you are buying | Which discipline | Our published floor |
|---|---|---|
| Evidence on where the real problem sits, in ten working days | Governance | Data Governance QuickScan from £14,500 + VAT |
| The operating model built on critical domains with the people who will run it | Governance | Data Governance Foundations from £45,000 + VAT over three to nine months |
| Senior data leadership without a permanent hire | Governance and direction | Virtual CDO from £7,500 + VAT per month |
| The governance function run for you month by month | Governance | Data governance as a service from £9,500 + VAT per month |
| A senior data lead inside a live programme | Both, at the join | Project DG from £22,000 + VAT per month |
| Target architecture for a data or AI platform | Management, decided properly | EA for data and AI platforms from £14,500 + VAT |
We do not quote platform licences, warehouse costs or engineering build rates, because we do not sell them and any figure we invented would be guesswork. What we will say is that UK data governance day rates commonly run from £1,100 to £2,400 per day, which is published on our own assessment page and is the reason we sell fixed fees instead. Our guide to data governance cost in the UK breaks the governance side down further.
The expensive version of this confusion has a familiar shape. An organisation buys a catalogue or a quality tool to solve a definitions problem. The tool works exactly as sold. Eighteen months later the catalogue has three competing entries for customer, because the missing thing was never software.
Regulation lands mostly on the governance side, because what regulators ask for is accountability and evidence rather than architecture. The ICO guidance on accountability and governance is explicit that being able to demonstrate compliance is part of the obligation, and demonstration requires named responsibility and records. A well engineered platform with nobody accountable for it does not answer that.
Quality sits at the join. The Government Data Quality Framework is a useful public reference precisely because it pairs measurement, which is management work, with ownership and accountability, which is governance work. The ONS material on quality in official statistics makes the same pairing from a different angle.
AI raises the stakes on the governance side specifically. A model inherits whatever the data already is, and the questions that stop AI projects are permission questions rather than pipeline questions. Whether this data may be used for this purpose, who decided, and on what basis. Our guides on data governance as the foundation of safe AI and the EU AI Act for UK companies cover where those obligations land.
We are a specialist advisory and assurance firm. Data governance sits alongside enterprise architecture, cyber assurance, ISO certification support and AI governance, which matters here because the governance and management split runs straight through all of them.
What we do. We establish where the real problem sits in ten working days, build the operating model with the people who will run it, provide senior data leadership on a fractional basis, run the governance function month by month where that is wanted, and set the target architecture for data and AI platforms so the management side is decided properly. Fees are published as floors and fixed in writing before work starts.
What we do not do. We are not a data engineering firm. We do not build pipelines, run warehouses, write transformation code or provide managed platform operations. We sell no software licences and take no vendor commissions, which is why our advice can end with a recommendation to buy nothing. If what you need is engineering delivery capacity, we will say so and we are happy to work alongside the firm that provides it.
If you recognise your organisation in the symptom table, the useful next step is short and fixed. Read the plain explanation of data governance if you want the ground beneath this article, the implementation framework if you have already decided, or the guide to choosing a data governance firm if you are comparing suppliers. When you are ready, speak to a senior practitioner and bring the two reports that disagree.
Data management is the work of moving, storing, modelling, protecting and serving data. Data governance is the work of deciding who may decide things about it, including definitions, ownership, access, quality thresholds and retention. The practical test is to ask what form the answer to a question takes. If the answer is a named person, it is governance. If the answer is a tool or a pipeline, it is management.
Recognised bodies of knowledge treat governance as one area within the wider discipline of data management, and that is defensible on paper. In practice they behave differently enough to be run separately, because one is a set of decision rights held by business people and the other is an engineering capability. Organisations that fold governance into an engineering function usually end up asking engineers to make business definitions, which does not work.
It depends on the symptom. Fund the platform first when it is genuinely unreliable, when a migration is already committed with a date, or when there is no shared data to govern yet. Fund governance first when numbers are disputed, an audit finding is open, an AI initiative is blocked on permission, two teams claim the same data, or nobody can say what may be deleted. If you cannot tell, a short assessment answers it with evidence.
No, although a tool makes governance easier to sustain once the decisions exist. Catalogues, quality platforms and lineage tools record and enforce decisions that people have already made. Bought before the decisions, they fill up with competing entries and stop being trusted. Ask who will maintain the contents, in whose working week, before signing.
Our governance work starts from £14,500 + VAT for a ten working day assessment, from £45,000 + VAT to build the operating model over three to nine months, from £7,500 + VAT per month for a Virtual CDO and from £9,500 + VAT per month for the governance function run for you. UK data governance day rates commonly run from £1,100 to £2,400 per day. We do not quote platform licences or engineering build costs because we do not sell them.
A named executive owns the operating model, and named business data owners own individual domains. Technology implements what those owners decide. Governance placed inside a technology function without business owners tends to stall, because the decisions it needs to make are commercial decisions about meaning and risk rather than technical ones.
UK data protection law requires organisations to be able to demonstrate compliance, which is why the ICO writes about accountability and governance together. Demonstrating it needs named responsibility, records of decisions and evidence that both are applied. That is governance work. A well built platform with nobody accountable for what it contains does not satisfy the requirement.
You need enough of both, focused on the data the AI will consume. On the governance side that means classification, a decision on permitted use and a named owner for each dataset in scope. On the management side it means measured quality on the specific fields the model reads. Neither needs to be complete across the whole estate before a first use case, and waiting for completeness is how AI programmes lose a year.
Emmanuel Olatunji
Chief Enterprise Architecture Partner
Emmanuel founded Intelance in 2017 and has led architecture, AI governance, cyber and deal engagements across the UK, Europe, the Middle East and Africa. Every Intelance engagement runs under his direction, and every guide he publishes comes from that work.
See his record on LinkedIn About IntelanceThirty minutes with a senior adviser. A clear view of where to start, what it costs and what your team can run themselves.
Talk to a senior adviser
Data Governance
Data Governance