In the insurance industry, fraud detection is no easy task, especially when you’re dealing with a company that has 5 million customers, served by 135 offices, 20,000 brokers, and 10,000 employees.

At Porto Seguro, the Fraud Detection unit needed a solution that could help analyze hundreds of thousands of proposals. The goal was to detect fraud more efficiently, preventing payouts on false claims.

The challenge included editing and creating rules, building risk scenarios, the ability to test those scenarios, and the ability to record processes for renewals or reopenings of future claims.

LOOKING FOR PATTERNS

In the system, the rules are atomic, and most involve a single variable. The main purpose of the rules is to score a fraud pattern (also called a scenario), using a given value to increase or decrease the score of that scenario. The rules also act as a precondition for new rules.

In addition to access to information about cases (accepted, rejected, borderline accepted, borderline rejected), there was an enrichment of the database with additional information for Business Intelligence analysis. It became possible to question which variables are (or are not) most frequent in fraud cases, for example. As a security measure, the system now prevents an employee from explaining to a broker the reason for a negative proposal response using sensitive data.

As an extension of the project, Stefanini Scala developed a multi-dimensional database for use with IBM Cognos in the calculation and delivery of specific indicators. These will support the discovery of new patterns for fraud detection.

Learn more by clicking here. Or talk to one of our experts!