Blog article
by Kamil Goral

Reducing False Positives In Fraud To 10%

Fraud prevention and loss recovery can be a cost-effective strategy to increase the bottom line, and with it comes a focus on minimizing the number of false positives in fraud.

False positives in fraud and the costs associated with them are one of the main challenges for people and teams working in the area of risk assessment. Regardless of whether we are talking about insurance false positives, procurement false positives or other risks related to crime, false alarm rates can reach into the 50-70% range, costing enterprises tens or even hundreds of millions of dollars a year.

One of the most important keys to fraud prevention is having the right fraud tool in place that enables you to minimize your exposure to risk while reducing your potential costs related with false positives in fraud.

Traditional Approaches

Most traditional risk assessment technologies operate on the Waterfall approach, consisting of various components for data integration (DI), data quality (DQ), etc. Projects are implemented in 6-24 months, and then after initial calibration by the vendor, often reach a false positives rate of less than 50%. So, the solution often starts out being relatively ineffective. Unfortunately, effectiveness tends to get worse over time as criminals change their methods, internal processes change and/or the internal structure of the customer portfolio changes.

Therefore, as shown in Figure 1 below, after a short time the number of false positives in fraud increases, and can reach a level of 70-80%. Very complex architectures, or alternatively closed black-box solutions, result in those solutions being difficult to implement, develop and maintain. Calibration requires the work of the vendor, and since these services typically are time consuming and expensive, they may be performed rarely or never.

Related Information

You Might Also Like

Fraud prevention and loss recovery can be a cost-effective strategy to increase the bottom line, and with it comes a focus on minimizing the number of false positives in fraud.

False positives in fraud and the costs associated with them are one of the main challenges for people and teams working in the area of risk assessment. Regardless of whether we are talking about insurance false positives, procurement false positives or other risks related to crime, false alarm rates can reach into the 50-70% range, costing enterprises tens or even hundreds of millions of dollars a year.

One of the most important keys to fraud prevention is having the right fraud tool in place that enables you to minimize your exposure to risk while reducing your potential costs related with false positives in fraud.

Traditional Approaches

Most traditional risk assessment technologies operate on the Waterfall approach, consisting of various components for data integration (DI), data quality (DQ), etc. Projects are implemented in 6-24 months, and then after initial calibration by the vendor, often reach a false positives rate of less than 50%. So, the solution often starts out being relatively ineffective. Unfortunately, effectiveness tends to get worse over time as criminals change their methods, internal processes change and/or the internal structure of the customer portfolio changes.

Therefore, as shown in Figure 1 below, after a short time the number of false positives in fraud increases, and can reach a level of 70-80%. Very complex architectures, or alternatively closed black-box solutions, result in those solutions being difficult to implement, develop and maintain. Calibration requires the work of the vendor, and since these services typically are time consuming and expensive, they may be performed rarely or never.

Figure 1. Black box & traditional approaches

The DataWalk Approach

DataWalk – reduction in insurance false positives

Figure 2. Continuous improvement with DataWalk enables a significant reduction in insurance false positives compared to alternative approaches

DataWalk is anti-fraud software for detecting and investigating potential frauds. DataWalk includes a powerful scoring engine which can be used for flagging suspicious claims, transactions, people, or anything else. As shown in Figure 2, DataWalk enables you to continuously improve your effectiveness in detecting fraud, ultimately enabling extremely high levels of accuracy and efficiency.

When designing the DataWalk system, we approached the issue of scoring from a different angle, knowing there are two keys to deliver superior performance in detecting suspicious conditions:

Fast results

With DataWalk we typically can install our software with the click of the button, import and connect your key data sets, rapidly customize and create rules and scores, and start getting first results in hours. Typically by day 7 you can already be discovering new patterns and fraud reduction opportunities.

You can then further improve performance by expanding your coverage using additional data sets, incorporating Machine Learning and DataWalk’s capabilities for Social Network Analysis (i.e., clustering) into your scores.

Do it yourself

DataWalk was designed to be self-service anti-fraud software. We provide a unique, flexible, powerful scoring engine that enables you to easily generate, monitor and continuously tune rules and scores yourself. You can easily test new hypotheses, add and test new data sources (e.g. OSINT - social media), generate new rules, change existing rules and weights, and deploy and manage machine learning models and other AI technologies. You can do all this yourself, very quickly with minimal effort and without professional services.

With this approach, you can continually improve your effectiveness in detecting more suspicious conditions, and dramatically reducing false positives in fraud. As summarized in Figure 2, other solutions start with mediocre performance that you can maintain only with regular incremental spending, while with DataWalk you can constantly improve your effectiveness as you continue to learn and adapt. DataWalk customers have used this approach of constant improvement to reduce false positive rates to 10% or less.

Reducing False Positives In Fraud To 10%

Related Information

Related Posts On The DataWalk Blog

You Might Also Like