Synectics Solutions have designed, piloted and deployed over 20 predictive models across 11 Tier 1 banks and insurance companies in the UK to help them improve their ability to prevent fraud, as well as substantially reducing the cost of doing so.
To help companies without access to sufficiently relevant target data Synectics has also built a standardised predictive analytics fraud
Optimise fraud prevention capability and reduce costs when launching new Current Account product.
Deployment of a standardised predictive analysis model in addition to SIRA fraud prevention solution.
Ability to identify almost 70% of fraudulent applications while only investigating around 15% of referrals.
Uplift in identification of fraud.
Reduced false positives.
Estimated savings of over £2 Million per year when product goes onto the market.
Synectics Solutions recently worked with a well-known financial brand to help them deploy a successful fraud prevention solution, despite the client having no relevant product historical data in the geography they were launching in.
The client wanted to ensure that when it launched its new Current Account it was optimising its fraud defences by utilising both the National SIRA fraud database, in conjunction with Synectics Precision predictive analytics capabilities. The following case study records the results of the proof of concept in preparation for the clients product launch.
Predictive analysis has become a widely used tool in financial services in the drive to improve fraud detection, and reduce investigation costs. However, many organisations struggle to deploy a viable predictive analytics programme because of a lack of sufficiently relevant or accurate target data, to build truly effective models.
Because of Synectics unique position, as custodian of the National SIRA fraud prevention database, the company used its data science capabilities, and ability to leverage the National SIRA database, to successfully build a standardised product specific predictive fraud prevention model to identify fraud - despite the client’s lack of sufficient in-house data.
Building the solution
Over 2 million historical current account applications were utilised to train the target model, which included adverse fraud data from National SIRA. Over 25 data features were then used to build the model, including personal applicant details along with additional data features only available from within National SIRA, such as historical adverse flags.
Once built Synectics comprehensively tested the model with the client in a proof of concept to prove its effectiveness.
The Precision Standard Current Account model was able to identify and prioritise fraud investigations to such a high degree that only 10% of referrals prioritised for investigation captured more than 50% of the adverse/fraudulent applications.
This represented a huge advantage for the client’s fraud team who would now be able to accurately reduce their investigative workload by a huge margin and still be certain of stopping most of the applications that were not just false positives.
Synectics estimates that this level of improved prioritisation and uplift in identification of fraud also achieved represents a likely financial benefit of around £2 Million over the course of the first year of deployment alone for the client.
Despite having no historical data footprint for this type of financial product we were able to use the Standardised Precision Current Account model to give us a major boost in identifying actual fraud, as well as improving efficiency and the ability to prioritise our fraud queues when our product goes onto the market”
Client Head of Financial Crime
“Some clients aren’t able to take advantage of our full enterprise Precision models due to various limitations such as lack of historical data, so it’s great to see our standard Precision models enabling clients like this to deploy otherwise unachievable day one fraud defences.”
Rob Bevington – Precision Product & Data Science Manager
PRECISION STANDARD – CURRENT ACCOUNT MODEL
Chart shows improvement in ability of Precision to prioritise and identify fraud beyond a rule-based only fraud solution as a result of the proof of concept pilot.
ROC CURVE FOR PRECISION PERFORMANCE
Not only was the Precision Standard model able to radically reduce the workflow of the client’s investigation team, but due to the additional model features deployed there were fraud’s that would not have been identified in a purely rule-based fraud work flow, leading to additional savings including;
Faster customer boarding – as good customers not held up in investigation queues.
Reduced workload for busy investigation teams.
Reduced False Positive Rates on investigations.
More effective identification of fraud than standard rules based solution alone.