Fraud Detection and Prevention Methods in Poland

Fraud detection is crucial for businesses operating in Poland. According to Marcin Zaboj, manager at KPMG’s actuarial services in Poland and CEE, identifying specific types of fraud may reduce the risk of that particular type of fraud, but it does not necessarily mean that overall fraud will decrease in the long term.

Assessing Risk of Fraud

To assess the risk of fraud efficiently, KPMG’s actuarial team in Poland has developed a comprehensive process for detecting potential fraud with an emphasis on its economic aspect. The process includes:

• Estimating the amount recovered as a result of identifying the fraud
• Analyzing the cost of verifying the fraud

Benford’s Law: A Tool for Fraud Detection

One approach to fraud analysis is based on searching for irregularities, including Benford’s law. This law describes the frequency distribution of the first digit in many real data sets and analyzes strong deviations from expected frequencies to identify suspicious or possibly manipulated data.

In Poland, Benford’s law can be used in state aid analysis, where there is a risk of data manipulation due to the intention of meeting certain requirements. Benford’s Law states that:

• Digit 1 occurs as the first digit 30.1% of the time
• Digit 9 occurs only 4.6% of the time

If manipulated or made-up numbers do not comply with Benford’s law, there is a risk that the data has been modified and requires more thorough analysis.

Fraud Detection Process

KPMG’s comprehensive process for detecting potential fraud includes:

Descriptive Analytics

• Defining what fraud means for the organization
• Supported by predictive analytics and big data

Predictive Analytics

• Identifying fraud patterns based on historical data collected within the organization and external data from a wider context
• Using machine learning methods to identify fraud patterns

Community Analytics

• Applying various tools to fully understand the characteristics and limitations of available data, including:
  • Handling missing values
  • Detecting and treating outliers
  • Defining flags
  • Standardizing data
  • Categorizing variables
  • Weighing evidence

Data Reduction

• Flagging damages depending on their characteristics
• Advanced analytics techniques used to build predictive models
• Searching for fraud patterns also includes social and geographical contexts, as the propensity to cheat is influenced by the social environment or geographic area

Model Assessment

The quality of the model is assessed by determining its prediction performance, and the stability of the data included in the model and the model’s calibration are tested. The whole process takes into account aspects specific to the organization, including current legislation such as the General Data Protection Regulation.

Conclusion

Advanced analytics play a key role in fraud management, and every organization should identify areas at risk of fraud and implement comprehensive fraud detection and verification processes to effectively respond to fraud challenges.