Case Study: Internal Fraud Analytics
A risk assessment scoring system for relationship managers
Successful on-going measurement of fraud events and activity allowing the audit department instant identification of potential fraudulent activity.
- Complex design requirement for a dynamically weighted scoring system to assess the risk at a resource level.
- KPIs needed to be scored from the CRM and multiple transactional systems.
- The transactional systems were distributed across 7 countries and had unique tabular designs.
- Test data was highly limited due to country specific private banking regulations and protection of account holder identification. Testing could only be incurred in-country, separate from country of development.
- The solution needed to be backward compatible – i.e. be able to test and score known past fraudulent activities.
- A relational database was used to build views of transactional and aggregated levels of data.
- A QlikView dashboard was designed to allow variable weighting of risk points, over 30 KPIs.
- Highly complex risk calculations were tested and validated in the front end before being moved to and optimized in the ETL process.
- The QlikView data model incorporated a linked table and dual calendar (data island) design.
- The audit department could instantly identify relationship managers that needed to be investigated internally.
- The ability to immediately change individual risk weightings allowed auditors to build a level of significance across their board of leading indicators.
- The client successfully back tested all known fraud events.
- The client identified multiple cases of potential historic fraudulent activity.
- Potential cases of active fraud where bough to the clients attention for immediate investigation.