Jaywing has been shortlisted alongside Virgin Money for Best Outsourcing & Partnership Initiative of the Year at the Collections & Vulnerability (C&V) Awards 2026.
The nomination recognises a joint project to develop a predictive machine learning model that not only identifies customers who may be at risk of attrition, but also estimates when that risk is most likely to occur.
Looking beyond the likelihood of attrition
For credit card providers, identifying customers who may leave is only part of the challenge. Many retention models provide a binary view of risk: a customer is either considered likely to leave or they are not.
What they don't show is when that risk is likely to materialise, making it difficult to know when to act. Intervene too early and a customer may not yet be showing meaningful signs of disengagement or wait too long and the opportunity to retain them may have passed.
Virgin Money wanted to understand that timing more clearly, so it could make retention activity more targeted and relevant.
One team, combining different expertise
Jaywing and Virgin Money worked closely throughout the project, bringing together their respective areas of expertise. Virgin Money contributed their understanding of the customer, business objectives and lending environment, while Jaywing brought its experience in machine learning and risk modelling. The resulting model was developed using Jaywing's Archetype platform and a Gradient Boosted Machine (GBM) framework.
Rather than creating separate models for different time horizons, the approach looks at attrition risk across six intervals: 3, 6, 9, 12, 15 and 18 months. This creates a forward-looking view of each customer's attrition risk, helping Virgin Money understand not just whether a customer may leave, but when that risk is likely to be highest.
Keeping machine learning explainable
Predictive performance is important, but so is understanding how a model reaches its conclusions, particularly in a regulated financial-services environment. Governance and explainability were therefore considered throughout the model development process. Constraints were applied to key variables to ensure that the relationships between risk drivers and model outputs remained consistent and logical. This helped create a model that combines strong predictive performance with the transparency needed to support effective risk governance.
Testing showed that the model achieved an 89.1% Gini coefficient for short-term attrition risk, while maintaining strong and stable predictive performance across all six-time intervals, extending to 18 months. Rigorous out-of-time validation also demonstrated the model’s stability, giving Virgin Money a more precise way to identify customers for targeted retention activity.
Nick Martin, Head of Lending Insight at Virgin Money, said:
“Our partnership with Jaywing has delivered more than just a model. It has delivered a new strategic capability. Being able to predict when a customer might leave, rather than simply whether they are at risk, was a true innovation born from a deeply collaborative process. It allows us to understand our customers on another level and engage with them at the most crucial moments. The result is a more proactive and customer-centric retention strategy that simply wouldn't have been possible without the close collaboration between our teams.”
Putting machine learning to practical use
AI and machine learning are increasingly being explored across financial services, but their value ultimately comes down to how effectively they solve real business problems. In this case, the focus was a specific one: giving Virgin Money a better understanding of when a customer may be at risk of attrition rather than simply identifying whether they are at risk.
The longitudinal approach provides a different way of looking at customer behaviour over time, while the model's governance framework helps ensure that the outputs remain understandable and usable within a financial-services environment.
For Jaywing and Virgin Money, the C&V Awards shortlist is recognition of what can be achieved when data science, customer insight and business expertise are brought together from the outset.