Three ways to tackle application fraud now
IFRS 9 – What you should you be thinking about now
How Jaywing's free syndicate and AI-powered modelling combats fraud
18 June 2019
Since 2001, we have worked with financial services providers to pioneer and perfect the use of pooled application data to help combat fraud. By collecting data from multiple lenders in syndicate, and using this to create highly accurate predictive models, lenders gain greater insight into possible fraudulent activity. Without this insight, lenders only have part of the picture, potentially exposing them to fraudulent applications and their associated losses. In this blog, Martin explores how recent changes to Foil could improve fraud detection by a further 10%.
Foil is Jaywing’s AI-powered fraud model solution. It uses pooled fraud outcome data from across the industry, through a data syndicate which is securely hosted by Jaywing. It’s free to join and share your data; working together, we can build better, more predictive models than any single organisation can achieve on their own.
How does Foil work?
Foil is a delivered as a bespoke application model, tailored for each portfolio, which delivers highly accurate fraud detection to lenders. Each model is based on a lender’s own application data, together with pooled fraud outcome data, ensuring lenders can minimise their exposure to fraud through the combined efforts of all the syndicate members.
Foil is powered by Artificial Intelligence, enabling lenders to better predict fraud using advanced, neural network-based techniques. Each model is delivered as a file of executable code, ready to be deployed quickly and simply within your existing data or decision environment. It needs no special technology to run.
Foil models enable you to prioritise those applications with the highest risk of fraud, typically enabling lenders to identify over 90% of fraud cases in the top-scoring 5-10% of their applications. This enables lenders to stem fraud losses as well as cutting fees for the use of third-party referral systems, massively reducing the cost of processing fraud.
By sharing anonymous samples of applications securely with Jaywing, including all cases of confirmed or suspected fraud, lenders can:
- Improve fraud detection - Take advantage of accurate application fraud scores based on industry-wide definitions of first and third party fraud.
- Tailor models to their own applicant profile
- Continue to protect customer data - Data is never shared with other lenders but is used anonymously within the modelling step to generate a more robust result to benefit all members.
- Prevent false positive results - While rules-based detection systems generate false positives, Foil’s models bring statistical accuracy into the process, homing in on the real combinations of features that are indicative of fraud.
How can Artificial Intelligence transform Fraud?
Foil is now underpinned by Archetype, Jaywing’s neural network-based modelling software. Foil’s AI-powered models are now over 10% better at identifying fraudulent cases than previous non-linear techniques, meaning that lenders can identify more fraud, with greater degrees of accuracy. What’s more, through our use of explainable AI we can help you delve into the model to enable you to understand the key drivers of fraud within your application process.
Not just available for consumer portfolios! We are also adding a focus on commercial lending and are looking for data sharing partners interested in stemming their losses in this growing area of fraudulent activity.
It’s free to join!
Joining the Foil syndicate benefits all lenders in their fight against fraud. It’s free to join.
As a member, Jaywing will host a copy of your fraud data and build a bespoke fraud model based on your own application data, using fraud data from across the industry as an outcome. We’ll share the model results with you and help you build a business case based on the benefits of deploying the model, comparing this with your current method of detecting fraud.
Although we are confident that Foil will generate significant savings in fraud losses and associated processing costs, there’s no obligation to go ahead and pay for the resulting model code, regardless of the ROI that we could realise.
How to get AI-powered fraud models
To access the AI-powered predictive model, there’s just a low, annual fee to pay to deploy the model within your own systems. This works in the same way that you currently deploy a credit score (alternative deployment options are available via an API).
The annual fee includes:
a) Delivery of model code in a language of your choice, for direct deployment within your systems
b) A yearly model refresh. Additional model refreshes are available for a small charge.