Data Governance: Small Rules, Ruthlessly Enforced
This is Article 3 of 4 in the Q-Lana Data and Knowledge Management series. Data governance works through a few small rules enforced without exception. It covers domain ownership, data quality rules, knowledge capture and where most implementations quietly give up.
Data governance is essential to reliable lending decisions, yet it is often treated as a set of policies, committees and standards that sit apart from daily work. The quality of a lender’s data is shaped by what people do every day, such as how a relationship manager records a client’s sector, how an analyst checks financial statements and how a credit officer records a covenant and its review date. If those tasks are inconsistent, the institution’s reports and decisions will reflect that inconsistency.
For SME and corporate lenders, data governance needs to be part of how work gets done. In this article, we look at how clear ownership, a small set of enforceable quality rules and better knowledge management can help institutions make more reliable decisions. The article follows the earlier articles on why SME lending transformations fail and the critical data domains for SME lenders, and builds on the data architecture discussed in Q-Lana’s minimum viable data architecture for SME lending.
Data governance starts with clear ownership

Data quality becomes difficult to manage when responsibility is unclear. Business teams may expect IT to maintain the information, while IT depends on staff in the business to enter it correctly and explain what it means. When nobody is accountable for resolving a problem, the same errors can appear in reports month after month.
Data governance roles: owners and stewards
Two data governance roles help address this. A data owner, who is a senior business leader accountable for a particular data domain and for decisions about its use. A data steward, who maintains definitions, monitors quality and helps teams apply agreed standards in their daily work.
The owner provides accountability, while the steward supports the work needed to maintain reliable information. Both roles need enough authority and support to resolve issues rather than simply document them.
Mechanisms that make ownership work
A small number of supporting mechanisms can make these responsibilities clear. Customer, facility and collateral identifiers should follow consistent naming conventions. Data quality service levels should set measurable tolerances, such as keeping missing sector codes below one per cent. Each exception should have a named owner and due date. Staff should also be able to trace a report, score or model input to its source and understand when it was last updated. Changes to important fields or definitions need an impact assessment and a rollback plan.
Fitting ownership into existing processes
Data governance arrangements should fit into existing lending and reporting processes wherever possible. If a rule requires extra work, the institution should be clear about the risk it controls and who is responsible for following it. Leadership support matters because staff will take data standards seriously only when managers apply them consistently, including when a transaction is urgent.
Data Governance: Small Rules, Ruthlessly Enforced
Many institutions already have data policies. The harder task is ensuring that important checks happen each time a record is created or updated. A manageable set of rules is usually a better starting point than a long catalogue that staff cannot apply consistently.
Five data quality dimensions for lenders
For lending institutions, five dimensions provide a useful basis for those checks.
- Completeness. Required fields such as legal form, sector code, beneficial ownership percentage, risk grade and collateral type should be completed before approval.
- Validity. Dates should make sense, interest rates should fall within approved limits and financial statement periods should follow a logical sequence.
- Accuracy. Information should be checked against reliable source documents. Beneficial owners, for example, should be verified against KYC records rather than copied from an unverified summary.
- Consistency. Each legal entity should have one active master record, while facility status and portfolio limits should agree with the relevant accounting and risk systems.
- Timeliness. Financial statements, covenant information and collateral valuations need to be updated within defined timeframes. For example, a lender may require statements to be no more than twelve months old at approval or renewal.
Owners, checkpoints and exceptions
The exact thresholds will depend on the institution, its products and its risk appetite. What matters is that each rule has a clear owner, a defined point at which it is checked and a process for dealing with exceptions. Where a rule is initially managed manually, the institution can automate it once it has been tested and shown to work as intended.
Data governance and BCBS 239
The Basel Committee stresses accurate, complete and timely risk information. This approach is consistent with its Principles for effective risk data aggregation and risk reporting (BCBS 239). The principles were developed with systemically important banks in mind, but they also provide useful reference points for institutions strengthening their own risk data practices.
Data governance programmes need sustained commitment

Data programmes can lose support when the early work is visible but the business benefits take longer to appear. Teams may have agreed definitions, assigned owners and produced a first quality scorecard, yet the scorecard may reveal that the data is less reliable than management assumed. At that point, leaders may question the investment before the new routines have had time to improve results.
Everyday exceptions erode standards
This is often when everyday exceptions begin to weaken the programme. A deal is urgent, therefore, a missing field is allowed through. An owner is not assigned because the relevant department is busy. A data issue is carried into the next reporting period. Each decision may seem reasonable in isolation, but repeated exceptions undermine the standards the institution has agreed to follow.
What should managers review?
Managers can reduce this risk by reviewing a small number of measures regularly, including missing-field rates, overdue updates, recurring exceptions and the time taken to resolve them. The review should lead to decisions about who will fix the problem and by when. It should also show whether the rules are reducing rework and improving the information used in credit reviews and portfolio monitoring.
Measuring the benefits
The benefits should be assessed in terms that matter to the institution. These may include fewer manual corrections, more consistent portfolio reports, quicker credit reviews and greater confidence in risk indicators. Progress will vary by institution, but regular measurement helps management distinguish between activity and improvements that are making a difference.
Data governance must preserve institutional knowledge

Structured data does not capture every factor that shapes a lending decision. Financial statements show performance over a period, but a relationship manager may also notice that a key customer has delayed payments, that management has changed or that a supplier relationship is becoming less reliable. Such observations can help explain emerging risks before they are reflected in the numbers.
Recording what staff observe
This knowledge is easily lost when it remains in emails, personal notes or the memory of an experienced employee. Lenders can reduce that risk by giving staff consistent ways to record site visits, reference calls and conversations with borrowers. The record should distinguish observed facts from professional opinions. Both can be valuable, but readers need to know which is which.
Linking notes to clients and decisions
The next step is to link these notes to the relevant client, facility, sector and risk assessment. Searchable records allow analysts and relationship managers to find earlier observations alongside financial information, rather than treating narrative notes as a separate archive. Credit committee rationales, exception approvals and post-mortems can also be retained so that future teams can understand why earlier decisions were made and what the institution learned from them.
Where AI can help
Natural language processing and generative AI can help summarize lengthy documents, identify recurring themes and connect narrative information to structured records. The usefulness of these tools depends on the quality and organization of the underlying material. AI cannot reliably recover institutional knowledge that was never recorded or resolve conflicting information without appropriate controls.
Trust is the currency

Lenders hold sensitive information, including financial statements, identity documents, ownership records and details about a borrower’s operations. Protecting it is part of maintaining client confidence and meeting legal and regulatory obligations. Access should be based on job responsibilities, and staff should see only the information they need to perform their roles.
Access, masking and retention
Personal information should be masked in test environments where possible, and data should be protected when stored and transmitted. Sharing with third parties needs a clear legal basis and documented controls. Institutions should also define how long records are retained, how they are disposed of and how access or changes can be traced during an investigation.
Building safeguards into procedures
These safeguards should be reflected in everyday procedures rather than left as statements in a policy. The European Banking Authority’s Guidelines on loan origination and monitoring provide a further reference for institutions reviewing governance across the credit lifecycle, including creditworthiness assessment and ongoing monitoring.
Data governance should improve lending decisions
Good governance is not measured by the number of policies an institution has approved. It is reflected in whether staff can rely on the information in front of them, whether a credit decision can be reconstructed later and whether important knowledge remains available when employees move on. These improvements also give analytical tools and future AI applications a more dependable basis for their work.
Keeping the work going
Data governance work is ongoing. Data owners and stewards need to review recurring problems, managers need to follow up on exceptions and staff need to understand why the rules matter. A short self-audit can help an institution assess its current position across completeness, validity, accuracy, consistency and timeliness, then identify the area that needs attention first.
Q-Lana’s Data & Knowledge Management for SME and Corporate Lenders white paper sets out the wider framework, including governance mechanisms and knowledge management templates. The article series also continues with a discussion of how AI uses the data and knowledge an institution has built.
All Articles in This Series
- Usable Data Is What Most Banks Are Missing
- Minimum Viable Data Architecture Rests on Eight Domains
- Data Governance: Small Rules, Ruthlessly Enforced (this article)
- AI in Lending Amplifies Whatever Data Is Already There
Follow Q-Lana
One short email a week on credit risk, lending and impact finance, free. Or browse more guides in the Q-Lana Knowledge Center.
Get Q-Lana Weekly
One short email a week on credit risk, lending and impact finance, free. Or browse more guides in the Q-Lana Knowledge Center.