Jaywing’s Proprietary Archetype Software Transforms Application Fraud Detection
Tue 26th Oct
- Risk consultancy works with lenders to generate predictive models for application fraud risk.
- Latest model generation achieves best-in-class results with predictive power of 97 per cent.
The market leading credit risk and data science consultancy Jaywing has announced a new service for lenders to transform the way they identify application fraud.
Previously, Jaywing led the way in developing explainable, controllable models for credit risk purposes through its AI-powered modelling platform, Archetype.
Applying the same platform with its deep learning techniques to fraud data, the consultancy is now able to exploit interactions between data points that are highly indicative of financial crime, generating models which predict the likelihood of fraud with unprecedented levels of accuracy.
Unlike a rules-based system or a traditional scorecard approach, Archetype’s models can identify pockets of application fraud which can only be detected through nuanced interactions between data in the account records and the data used to validate them. Typically, the models use application information, bureau data, and any fraud referral rules generated from external data sources such as CIFAS, Hunter, Detect, etc, that the lender already uses. As the model consumes these rules, the lender doesn’t need to review individual rule triggers separately.
Archetype results have proven to be more than twice as effective as the incumbent fraud approach. For example, a model built for a major UK high street bank has achieved a Gini approaching 97 per cent (Gini being a measure of the predictive power of the model). What this means in practice is that over 88 per cent of attempted application fraud was contained within the top-scoring five per cent of applications. This means that the lender can minimise fraud declines with the expectation of exceptionally low fraud losses.
For a credit card company, Archetype has demonstrated the power of AI scoring adoption through a model which will deliver a dramatic cut in manual review and a reduction in third-party data costs without any reduction in fraud detection levels, using a model which achieves a similarly impressive Gini of 93 per cent and an annual benefit of at least £1.5 million per year.
This means lenders do not need to invest in any special technology in order to use the models, which can be deployed in any modern credit decisioning system or accessed via an API call. Archetype uses the same data inputs that lender already uses to identify credit risk and application fraud, retaining the same data calls and services as are currently in force.
Ben O’Brien, Managing Director of Jaywing’s Financial and Professional Services Team commented, “We’ve been fans of using non-linear techniques to detect fraud for many years, but using Archetype brings the full predictive power of neural networks to bear on the application fraud detection problem – with incredible results. Archetype squeezes every drop of predictive power from data you already collect, reducing fraud referrals without increasing losses. It puts a laser focus on where your suspect cases are likely to be, making your fraud process as efficient as possible.”
For more information on how our industry-leading bespoke modelling approach can transform your fraud detection, contact Glen Graham at firstname.lastname@example.org