Completed Project Case Study: Improving Financial Inclusion through Data-Driven Credit Union Analytics 

The Challenge 

Financial exclusion remains a major challenge in the UK, affecting millions of people who struggle to access affordable credit, build savings, or improve their financial security. People with irregular incomes, limited credit histories, or previous repayment difficulties can find themselves excluded from mainstream financial services and reliant on high-cost borrowing. 

Credit unions play an important role in providing fair and affordable financial services to underserved communities. However, many credit unions have limited capacity to analyse complex customer data, identify emerging signs of financial distress, or tailor their support to different customer needs. 

This project set out to explore how mathematics, data science, and responsible machine learning could help a credit union better understand borrowing and repayment behaviour, strengthen responsible lending, and intervene earlier when members begin to experience financial difficulty. 

The Project 

Led by Professor Olivier Menoukeu Pamen from the Department of Mathematical Sciences at the University of Liverpool, the project was delivered in partnership with Enterprise Credit Union (ECU), a Liverpool-based credit union serving communities across Merseyside. 

The research team analysed five years of anonymised monthly loan and repayment data covering the period from 2020 to 2024. The data included information on customer borrowing, repayments, refinancing activity, and account status, allowing the researchers to track how financial behaviour changed over time. 

The project combined data engineering, behavioural analysis, statistical modelling, and machine learning. It aimed not only to understand financial risk, but also to identify when and how targeted support might prevent financial distress and help members build stronger credit records. 

The team cleaned and restructured ECU’s customer and loan data, analysed borrowing, repayment, refinancing and rollover patterns, and grouped members into distinct behavioural profiles. It also identified how repayment difficulties developed over time, built predictive models to detect early signs of delinquency and recovery, and translated the findings into practical recommendations for lending, collections and member support. 

Key Outcomes 

A major outcome of the project was the improvement of ECU’s data foundations. The research team created a cleaner, more structured, and decision-ready data environment that improves how loans, repayments, and refinancing activity are recorded and analysed. 

The project also introduced a more behaviourally accurate way of identifying new loans. Traditional account-based measures can understate borrowing activity when customers refinance or roll over existing balances. By distinguishing between account records and actual borrowing behaviour, the project provided ECU with a clearer picture of how often members were borrowing and how exposure developed over time. 

The analysis showed that repeated borrowing and balance rollovers were common among ECU members, reflecting income volatility and limited access to mainstream financial services. Although most repayments were on track in any given month, financial stress was concentrated among a smaller group of members with recurring patterns of difficulty. This was an important finding because a generally stable loan portfolio can conceal persistent vulnerability among particular customers. Without earlier support, these members may move into repeated refinancing, growing balances, and longer-term debt cycles. 

The project identified clear behavioural pathways through which repayment difficulties emerge. Early rollover stages were more unstable, while later stages became more predictable but were associated with higher accumulated exposure. The analysis also identified moderate delinquency as a key tipping point, after which repayment difficulties were much more likely to escalate. Once severe problems occurred, recovery became considerably more difficult without targeted intervention. 

A central output was a behavioural persona framework that grouped members into seven clear profiles, ranging from stable, lower-risk borrowers to customers experiencing severe financial distress. Each profile was linked to potential approaches for credit decisions, collections, product design, financial education, credit-building, and hardship support. 

Collaboration 

The collaboration with Enterprise Credit Union was central to the success of the project. ECU staff worked closely with the research team to explain the organisation’s data, lending processes, operational priorities, and customer context. 

Regular meetings allowed the partners to resolve questions early and ensured that the research remained grounded in ECU’s day-to-day experience. A key lesson was that data definitions and operational terms used by practitioners can differ significantly from academic interpretations. Establishing a shared understanding of these concepts was therefore essential before the analysis could progress. 

ECU also allocated a dedicated data analyst to support the work and provide timely clarification. Where broader strategic input was required, the Head of Operations contributed to discussions. This structured engagement reduced delays, built trust, and enabled the research team to develop outputs that were relevant to both operational staff and senior management. 

The collaboration demonstrated the value of continuous dialogue and co-development. Rather than producing an academic model in isolation, the team worked with ECU to ensure the findings reflected the realities of responsible lending within a community finance organisation. 

Impact 

The project has provided ECU with a clearer and more evidence-based understanding of its loan portfolio and customer behaviour. The findings helped confirm and quantify patterns that staff had observed through experience, while also identifying new opportunities for earlier intervention. 

The behavioural profiles and predictive models could help ECU recognise when a member is beginning to experience financial difficulty and respond before the situation becomes severe. This creates opportunities for more proportionate and supportive action, such as tailored financial education, revised repayment arrangements, credit-building pathways, or specialist hardship support. 

The project also has the potential to influence ECU’s internal policies and operational practices. The findings have been presented to the Chief Executive Officer and have contributed to early discussions about credit risk assessment, collections, early intervention, and the future use of behavioural analytics in responsible lending. 

Although no formal policy changes have been introduced to date, the engagement with senior leadership has created a clear route for the research to inform future decision-making. The team has also been invited to present its findings to ECU credit officers, providing an opportunity to test the practical relevance of the framework with frontline decision-makers. 

Beyond ECU, the project demonstrates how transparent and explainable data models can provide an alternative to opaque, one-size-fits-all commercial credit scoring systems. This is particularly important when working with financially vulnerable customers, where poorly designed models could reinforce existing inequalities. 

The research therefore provides a potential blueprint for how other credit unions could use data science and mathematical modelling to improve responsible lending, identify financial vulnerability, and support the long-term wellbeing of their members. 

Going Forward 

The immediate next step is for the project team to present and discuss the findings at a workshop with ECU credit officers. This will help translate the analytical results into practical lending and collections activity and provide feedback from staff who work directly with members. 

There is also scope for staff training in Python, statistical analysis, and machine learning. This would enable ECU’s data analyst and other relevant staff to interact directly with the models, reproduce the analysis, and adapt the tools as operational needs change. 

ECU has also expressed willingness to host the Research Assistant who worked on the project for a period of implementation, training, and knowledge exchange. Although no formal placement has yet been arranged, this could provide a valuable route for embedding the research within ECU and maintaining continuity between analysis and practical adoption. 

UKFin+ Role 

UKFin+ funding enabled the University of Liverpool and Enterprise Credit Union to explore how advanced data analytics could address a real-world financial inclusion challenge. 

The funding created the space for the partners to improve ECU’s data infrastructure, undertake detailed behavioural analysis, test responsible machine learning approaches, and translate academic research into practical tools for a community finance provider. 

By bringing together mathematical expertise and the operational knowledge of a credit union, UKFin+ helped establish a longer-term collaboration with the potential to strengthen responsible lending, improve support for financially vulnerable members, and provide a model that could be adopted more widely across the UK credit union sector.