An Insurance Giant Calls on Talan to Reduce Auto Insurance Fraud
Context
Our client is an international player in insurance and assistance, present in 64 countries and serving nearly 107 million customers worldwide with a turnover of €24.5 billion. In France, this represents:
6,3 million
customers
33 000
employees
thousands
of partner dealerships
Challenges
The company called on Talan to fight auto insurance fraud and evaluate the contribution of machine learning for fraud detection. The objective was to reduce auto insurance fraud originating from dealerships without altering the customer experience.
Methods
Talan teams used Machine Learning to identify new high-impact rules that are easily integrated into the existing system, and to industrialize the model(s) using Spark—a unified, ultra-fast analytics engine for large-scale data processing.
In a second phase, Talan teams built a fraud management dashboard using the TIBCO Spotfire analytical solution, allowing for real-time tracking of alerts and investigation results.
Benefits
A successful experiment that allowed for the detection of 30% additional fraud without increasing the number of false positives
The number of investigated invoices was multiplied by two
No negative impact on the customer experience
A new algorithm specifically developed by Talan, generating rules that are directly ready for industrialization