Background
Increasing digitization of services offered; changes in bank consumption behaviors; development of complex fraud schemes…traditional anti-fraud controls have reached the limits of their effectiveness: they can only find what they are looking for and generate a growing volume of false positives, which take up analysts’ time. In such an environment, a private bank gives itself flexibility to better protect its customers, better meet the regulatory requirements, and manage the workload.
The client and Wavestone chose to construct an effective, projective response based on Machine Learning.
Challenges
In this approach, the client and Wavestone responded to three challenges:
- Revealing clues in addition to alerts.
- Developing a self-learning radar for operations and behaviors specific to each customer without a pre-established pattern.
- Making the anti-fraud system more data-driven: developing expertise in analyzing indications, re-implementing resources, reviewing procedures, etc.
Responses & Key Success Factors
Alongside the client, Wavestone designed, prepared, and expanded the system for early detection of differing behaviors:
- Effective, projective approach: gradual introduction of levels of earnings, controlled commitments of resources, and re-examination of trajectory based on the results achieved.
- Development of the self-learning radar: innovative Machine Learning application combining different types of learning and supported by Wavestone’s R&D work. Connection to the core banking system.
- Start-up of the indication analysis unit.
