Why Data Governance Programmes Lose Momentum After Launch
Most governance programmes do not collapse. They fade. The causes are structural, they are visible early, and each has a different remedy.
A data governance programme can end without anyone deciding to stop it. It ends by degrees. The council that met monthly meets quarterly, and then when someone remembers. Senior members send deputies. Stewards stop raising issues because the last ones went nowhere. The policies remain on the intranet, accurate and unread. A year after launch the programme still exists on the organisation chart and has stopped changing anything.
If that describes your programme, the instinct may be to relaunch it with a new sponsor, a refreshed framework and another round of communications. That treats the problem as lost enthusiasm. The patterns below point to structure instead: programmes fade because of how they were built. They were given meetings and no decisions to take, roles that carried neither authority nor time, and policies that had no route into anyone’s work. A second launch on the same design fails in the same way.
This article sets out eight patterns behind the fade. For each it describes what you would see, what usually lies behind it, a question that tells you whether it applies, what to do, and what the remedy costs.
What the evidence does and does not show
A note on sources first, because this subject attracts confident numbers.
Articles on this subject commonly open with a failure rate. In the research for this article we found no reliable published measurement of how often data governance programmes fail. The figures we examined could not be verified against a primary source to the standard this article requires, so none is used here and we do not present a universal failure rate. That is a statement about the evidence we identified and verified. It is not a claim that no such evidence exists.
What does exist is detailed public evidence about why governance of data stalls in particular settings:
- The Basel Committee on Banking Supervision’s 2023 assessment of how 31 of the world’s largest banks had adopted its principles for risk data, nearly ten years after they were issued1.
- The UK National Audit Office’s 2019 examination of how central government uses data2.
- Peer-reviewed research on why governing data proved difficult across 13 Danish municipalities3.
These are banks, government departments and local authorities. None of them describes organisations in general, and the bank evidence concerns risk data specifically. The patterns below are our analysis of what these sources have in common. The diagnoses and remedies are practitioner analysis and should be read as that. They are not findings of the sources, and they are not drawn from client engagements.
Eight patterns behind the fade
1. The forum meets but does not decide
What you see. Agendas are made of updates. Papers are “noted”. The same items return each month. Attendance drifts from principals to deputies, and then lapses.
What is usually behind it. The forum was set up as a structure before anyone settled what it was for. Governance is, at root, a matter of which decisions must be made and who is accountable for making them4. A forum with no decisions assigned to it has nothing to do except receive reports, and senior people soon notice that their presence changes nothing. The National Audit Office recorded the end state in UK central government: the two cross-government groups on data had not met regularly in 20192.
The question to ask. In the last three meetings, what did this forum decide that would not have been decided without it?
What to do. Write down what the forum is entitled to approve, what it must refer upward, and what is delegated below it. Rebuild the agenda from decisions that are waiting. If there are too few, meet less often. A meeting that takes one real decision a quarter is worth more to its members than twelve that take none.
The trade-off. Real decision rights take authority from somewhere else, which may mean from functional heads who have been deciding informally. That conversation is uncomfortable and cannot be avoided. Without it the forum’s authority remains nominal.
2. Sponsorship that was a signature, not attention
What you see. A senior sponsor approved the programme and has not been involved since. Budget requests stall. When governance and a delivery deadline conflict, the deadline wins and nobody intervenes.
What is usually behind it. The sponsor agreed to the idea of governance and was never asked to do anything specific. The Basel Committee’s findings are direct on this point. Banks’ adoption programmes “were often underfunded, limited in scope and lacking sufficient attention from boards of directors (boards) and senior management”, and at certain banks leaders lacked awareness of data issues “and therefore do not ensure appropriate budget, resources and accountability”1. The Committee attributed delayed compliance largely to lack of prioritisation and insufficient ownership at the top.
The question to ask. What has the sponsor personally decided, funded or unblocked in the past six months?
What to do. Give the sponsor a small number of defined acts: resolving an escalated dispute between two functions, approving a policy with real consequences, reviewing a short set of outcome measures each quarter. Sponsorship is sustained by having a job to do. If the sponsor will not take on even that, the programme is under-sponsored, and it is better to know.
The trade-off. Asking more of a sponsor risks being told that governance is not a priority. That answer is useful. It says that the scope should shrink to what a less senior sponsor can carry.
3. Stewards with a title and no time or authority
What you see. Stewards were named at launch. Their managers were not consulted about the time involved. Stewardship work is done late or not at all, and stewards who try to enforce a standard find that they have no standing to do so.
What is usually behind it. Stewardship was added to existing jobs as an unfunded extra. The research on Danish municipalities frames governing data as a collective action problem, and names capabilities, collaboration and organisational politics among its challenges3. In a problem of that kind, the people asked to contribute effort are not necessarily those who benefit. The National Audit Office made the practical version of the point in its recommendations: guidance has to “recognise the effort and resource required to fully and consistently adopt the policy and principles” in working practice2.
The question to ask. Is stewardship written into anyone’s objectives, with time allocated, and does the steward’s line manager know?
What to do. Reduce the number of stewards and make the role real for those who remain: stated time, a place in their objectives, and a named owner with authority standing behind them. Be explicit about what a steward may decide alone and what goes to the owner. The roles and how to assign them are covered in a separate guide.
The trade-off. Fewer, properly resourced stewards means less data is covered at first. Nominal coverage is of little use, so a smaller estate that is actually governed is the better starting point.
4. Scope that is too wide to finish or too narrow to matter
What you see. Either the programme set out to govern all enterprise data and, two years on, is still compiling inventories; or it was confined to a single compliance requirement and the rest of the organisation regards it as someone else’s concern.
What is usually behind it. Scope was set by ambition or by a regulation, with too little regard for what the organisation could absorb. The Basel evidence shows the failure in both directions. Programmes were “limited in scope”, and banks that treated the principles narrowly could regard themselves as compliant while they continued “to struggle with systemic data quality issues”. At the same time, some banks expanded the scope of their plans as their businesses changed1. Nor is there a standard scope to copy. The Basel Committee accepts that each bank may have a governance arrangement that best fits its mix of business activities and organisational structure1, and Khatri and Brown conclude that different arrangements can suit different areas of decision within the same organisation4.
The question to ask. Can you name the specific data this programme governs today, and would the people who depend on that data say it has improved?
What to do. Define the scope as a list of data that matters, chosen by business consequence, and govern that completely before widening it. Starting with critical data elements sets out one way to draw the boundary.
The trade-off. A narrow scope is easier to finish and easier to dismiss. Choose data that a senior leader visibly cares about, so that the first result is one the organisation notices.
5. Policy published, practice unchanged
What you see. A full set of policies exists. Few people outside the governance team could say what they require. Projects go live without reference to them, and nothing happens as a result.
What is usually behind it. Writing policy was treated as the deliverable. Policy changes behaviour only where it meets a process people already follow, such as a project gate, an access request or a change procedure. Regulators take the same view of formal compliance. The UK Information Commissioner’s Office treats accountability as a continuing obligation that has to be reviewed and kept up, not a state reached by signing a document5.
The question to ask. For each policy, at what point in someone’s ordinary work would they be stopped if they did not comply?
What to do. Pick the two or three policies that carry real risk and attach each to an existing control point. Shorten them until a non-specialist can apply them. Retire or suspend the rest openly, because a policy that is known to be ignored weakens the ones that matter. Governing self-service BI is a useful test case, because the policy has to be built into the tooling to work at all.
The trade-off. Enforcement at a control point slows delivery, at least at first, and that will be raised as an objection. Have the answer ready: what the control prevents, and how long it actually takes.
6. No connection to what the organisation is trying to deliver
What you see. Governance has its own plan, its own milestones and its own vocabulary. Delivery programmes regard it as a separate workstream that asks for things. When priorities tighten, governance staff are reassigned.
What is usually behind it. The programme was justified in general terms, such as better data, reduced risk or a single version of the truth, and never attached to an outcome that someone senior is measured on. The National Audit Office observed what follows. The department leading data policy had not made the progress it expected, largely because its staff were diverted to work on EU exit2. The research on municipalities names “value” among its six challenges: the benefit is hard to demonstrate to those asked to pay for it3.
The question to ask. Which funded business initiative would be at risk this year if the governance programme stopped tomorrow?
What to do. Find the initiative and work for it. A regulatory deadline, a platform migration, a customer programme or an AI use case each depends on particular data being owned, defined and fit for purpose. Make that the governance work plan. The National Audit Office’s recommendation that departments set out data requirements in business cases points the same way2: governance should enter through the investment process and not sit beside it. The parallel problem in strategy is examined in why data strategies stall after approval.
The trade-off. Tying governance to one initiative makes it dependent on that initiative’s fortunes, and it will leave gaps elsewhere. Accept the gaps. A programme that is needed by something survives. One with complete coverage that nothing depends on is the kind that fades.
7. Issues accumulate and people work around them
What you see. The issue log grows. Items stay open for months without an owner. Analysts keep private corrections in spreadsheets. New issues stop being raised, which can look like improvement.
What is usually behind it. There is a way to raise issues and no mechanism that obliges anyone to resolve them, so the log has become an archive. People then conclude, reasonably, that reporting a problem achieves nothing, and they fix it locally. The National Audit Office named the result: “a culture of tolerating and working around poor-quality data”2. Once that culture is established it hides the problem, because the visible signal, the number of new issues, falls.
The question to ask. How old is the oldest open issue, who owns it, and when was it last discussed by someone able to resolve it?
What to do. Give every issue an owner and a severity based on business consequence. Set a limit on how long an issue may stay open at each severity before it is escalated automatically. Close or formally accept the old ones, and say which. An explicit decision to live with a defect is a governance outcome, and a better one than leaving it open without a decision.
The trade-off. Cleaning the log will expose how much has been tolerated, and for a time the picture will look worse. Tell the sponsor in advance that this will happen.
8. Measures that count activity
What you see. Reports to leadership list policies approved, stewards appointed, terms defined and meetings held. All the indicators are green. The people who use the data report no difference.
What is usually behind it. Activity is easy to count and outcomes are not, so the programme reports what it can. The risk is that it comes to believe its own reporting. The Basel Committee found that banks tended to rate their own adoption more favourably than their supervisors did, and described a “dichotomy” in which certain banks considered themselves fully compliant while multi-year remediation of systemic data quality problems was still under way1.
The question to ask. If the programme’s dashboard were shown to the heaviest users of the data, would they recognise it as a description of their experience?
What to do. Replace most activity counts with a few outcome measures: whether priority data meets its stated quality expectations, how long issues take to close, how often key reports have to be manually corrected. The Basel Committee’s recommendation to banks is a sound model: present the board with a standard set of indicators that allows it to assess data quality for itself1. Where possible, have someone outside the governance team verify the numbers.
The trade-off. Outcome measures will show less progress than activity measures did, and they depend partly on things the governance team does not control. That is uncomfortable, but leaders can only act on measures that describe what users experience.
Working out which pattern you have
These patterns overlap, and a fading programme may show several. They are not equally fundamental. The order below runs from causes to consequences, and it is the order in which to check.
Diagnosing a fading governance programme: where to look first
| Check | Pattern | Earliest visible sign |
|---|---|---|
| 1 | No connection to delivery priorities | Governance staff are the first to be reassigned |
| 2 | Sponsorship without attention | Escalations to the sponsor go unanswered |
| 3 | Scope too wide or too narrow | Nobody can list the data actually governed |
| 4 | The forum does not decide | Agendas of updates; deputies attending |
| 5 | Stewards without time or authority | Stewardship tasks slip quietly |
| 6 | Policy without adoption | Projects go live without reference to policy |
| 7 | Issues accumulate | Old open issues; fewer new ones raised |
| 8 | Measures count activity | Green reporting that users do not recognise |
The first three are conditions of the programme’s existence. If governance is attached to nothing the organisation needs, has no active sponsor and has no defined scope, repairing the forum or the issue log will not hold. The last five are mechanisms, and they can be fixed one at a time once the first three are sound.
Repair or reset
Where a fading programme can be repaired in place, that is the better course. A reset discards relationships and credibility that took time to build.
A reset is the honest choice in two situations. The first is when there is no business initiative that needs the programme and no sponsor prepared to act. In that case the kindest step is to shrink governance to the minimum the organisation is obliged to maintain, state plainly that this has been done, and wait for a real need. The second is when the programme’s name has become a liability, so that anything carrying it is discounted. Then the substance can continue under the initiative it serves, without the branding.
Two limits should be kept in view either way. Governance is not a substitute for investment: the Basel Committee noted that a strong governance framework alone cannot overcome legacy systems, manual processes and fragmented data1. And the evidence does not support a promise that any of these remedies will work in a given organisation. The sources establish where programmes have fallen short. They do not test the fixes.
Where to go next
If you want a structured view of where your programme stands before deciding what to repair, the Governance Maturity Assessment covers the same ground as the eight patterns in assessment form. Use it to find the weakest of the first three conditions. A score on its own tells you little.
For the underlying model, see what data governance is and what a working programme contains. To rebuild accountability, see data owners, stewards and custodians. To narrow the scope, see starting with critical data elements.
Sources
- Basel Committee on Banking Supervision (Bank for International Settlements). Progress in adopting the Principles for effective risk data aggregation and risk reporting. 28 November 2023.
- National Audit Office (UK). Challenges in using data across government (Report by the Comptroller and Auditor General, HC 2220, Session 2017-2019). 21 June 2019.
- Benfeldt, O., Persson, J. S. and Madsen, S. Data Governance as a Collective Action Problem. Information Systems Frontiers, 22(2), 299-313. 2020.
- Khatri, V. and Brown, C. V. Designing data governance. Communications of the ACM, 53(1), 148-152. 2010.
- Information Commissioner’s Office (UK). Guide to accountability and governance (UK GDPR guidance); UK GDPR Article 5(2).
This article is analysis of published sources. It is general professional information, not legal or regulatory advice, and it does not describe any specific organisation’s programme beyond what the cited sources report.