By Artem Lalaiants
European lenders face a difficult contradiction.
They need better information to assess applicants. Yet every new data source creates another layer of regulatory, operational, and reputational risk for credit decisioning software.
Traditional credit files remain essential. But they do not always show the full picture.
A young professional may have stable income but little borrowing history. A recent migrant may have years of responsible financial behaviour outside the local bureau system. A gig worker may earn consistently without fitting standard employment models.
Alternative data can help lenders understand these applicants. Open banking records, rent payments, utility history, telecom data, and digital identity signals can add useful context.
Europe, however, does not treat more data as automatically better. Regulators ask harder questions. Is the information relevant? Is its use proportionate? Can the decision be explained? Could the model disadvantage certain groups?
This approach is often described as a barrier to innovation. I see it differently.
Europe is creating a market where trust, governance, and explainability will determine which credit technologies succeed. For business leaders, that raises the standard. It also creates a more sustainable path to innovation.
Better predictions are no longer enough
Credit risk teams already operate under intense pressure.
They must approve more creditworthy applicants without increasing defaults. They must also prevent over-indebtedness, reduce fraud, protect personal data, and explain important decisions.
A small error can affect thousands of applicants. It can also lead to regulatory action, financial losses, and lasting reputational damage. Risk teams understand these consequences better than anyone.
Alternative data adds another challenge. It expands the evidence available for assessment, but it also creates new questions about relevance and fairness.
Some signals have a clear connection to financial behaviour. Open banking data can show income, expenses, and existing commitments. Rent and utility payments can provide evidence of regular payment patterns.
Digital identity data can add a different type of context. It may show whether an email address has existed for years, whether applicant details remain consistent across sources, or whether a phone number and device appear stable.
Other signals require much more caution. Their relationship with credit risk may be weak, indirect, or difficult to justify. The fact that information exists does not mean it belongs in a credit model.
This changes the strategic question.
Lenders should not ask, “What data can we access?” They should ask, “What evidence do we need to make this decision responsibly?”
Automated decision-making creates another layer of scrutiny. GDPR Article 22 restricts certain decisions based solely on automated processing when they produce legal or similarly significant effects. It also supports safeguards such as human intervention and the ability to contest a decision.
The recent SCHUFA judgment strengthened that principle. The Court of Justice of the European Union found that creating a credit score may itself qualify as automated decision-making when a lender relies heavily on it.
Responsibility therefore does not begin only when a lender approves or rejects an application. It can also sit within the systems that produce the recommendation.
The EU AI Act pushes this further. AI systems used to evaluate an individual’s creditworthiness or establish a credit score are generally treated as high-risk. These systems face stricter expectations around data quality, risk management, documentation, transparency, monitoring, and human oversight.
For credit leaders, the message is clear. Predictive performance remains essential, but it is only one part of model quality.
Building governance into the scoring strategy
Compliance cannot sit beside the model as a separate legal exercise. It must shape how teams select data, build systems, and use their outputs.
In my experience, three principles make this work more practical.
1. Prove that every data category belongs
Alternative data should answer a specific risk question. It should not enter a model simply because it is available.
Teams should document why each data category matters. They should test whether it improves decisions and whether a less intrusive source could achieve the same result.
This is particularly important for digital footprint analysis. Useful digital signals may help assess identity consistency, account maturity, behavioural stability, or potential fraud. But providers must separate signals with a defensible risk relationship from digital activity that is merely interesting.
Data minimisation is not only a privacy requirement. It can improve model discipline. Removing irrelevant information reduces noise and makes the final assessment easier to explain.
2. Design explanations for real decisions
Explainability should not mean giving a consumer a technical description of an algorithm.
Risk teams need reasons they can use. They should understand which factors influenced the output, how those factors affected the assessment, and what limitations apply.
A lender should also be able to investigate an unusual result. Human reviewers need enough context to challenge a score rather than simply approve it.
A model becomes more useful when teams can act on its reasoning. Explainability therefore supports both compliance and better operations.
3. Treat fairness as continuous model work
No dataset remains representative forever. Customer behaviour changes. Products enter new markets. Economic conditions shift.
Bias testing cannot remain a one-time approval step before deployment. Teams need regular monitoring across relevant customer groups and lending outcomes.
They should track more than model accuracy. Approval patterns, manual overrides, complaints, data errors, and unexpected score movements may expose risks that standard performance measures miss.
When teams find an uneven outcome, they need a clear investigation process. The issue may come from the source data, a model feature, a lending policy, or the way employees interpret the output.
Share responsibility across the credit ecosystem
Modern credit decisions rarely depend on one organisation.
A lender may use information from credit bureaus, open banking providers, fraud platforms, alternative data companies, and internal customer records. Each party controls a different part of the process.
This makes responsibility easy to fragment.
A technology provider may argue that the lender makes the final decision. The lender may assume the provider has already tested every feature and data source. Neither assumption gives risk teams enough protection.
Contracts should clearly define who manages:
- data quality and accuracy;
- model testing and ongoing monitoring;
- consumer access, correction, and review requests;
- security incidents and regulatory reporting;
- documentation for audits and supervisory reviews.
Providers should give lenders enough information to understand and supervise the system.
Lenders must still examine how the output works within their own credit policies. The same score can produce very different outcomes depending on cut-offs, loan terms, target segments, and review procedures.
Governance must therefore cover the full decision chain, not only the model itself.
This shared approach can reduce pressure on internal risk teams. Responsible providers should support them with clear documentation, transparent reason codes, audit trails, and reliable performance monitoring.
Turning regulatory pressure into long-term advantage
Europe’s rules raise the standard for credit innovation. They also make poorly governed experimentation harder.
That can benefit lenders that invest early in trustworthy infrastructure.
A well-governed alternative data strategy can help teams assess more applicants without abandoning established controls. It can support thin-file decisions, strengthen bureau models, and provide more evidence when traditional information leaves uncertainty.
The competitive advantage will not come from avoiding regulation. The only right way is to learn to operate confidently within it.
Leaders should look for systems that combine predictive value with clear governance. Data relevance, fairness testing, human review, and explainability should not be optional additions.
Europe is not forcing lenders to choose between financial inclusion and responsible risk management. It is asking our industry to prove that the two can work together.
The next generation of credit scoring will not be defined by how much technology it uses. It will be defined by how well people can understand, govern, and trust the decisions that technology supports.
About the Author
Artem Lalaiants is the CEO and Co-Founder of RiskSeal, a US-based startup developing digital identity verification and alternative credit scoring technology. He has led the company from concept to adoption by digital banks and consumer lenders, helping risk teams use real-time digital footprint signals to support faster and more informed credit decisions.
Disclaimer: This article contains sponsored marketing content. It is intended for promotional purposes and should not be considered as an endorsement or recommendation by our website. Readers are encouraged to conduct their own research and exercise their own judgment before making any decisions based on the information provided in this article.
