The key to reducing false positives in fraud detection systems

For the uninitiated, if we go back a few years the term ‘false positive’ was mainly a medical phrase that indicated that a patient had been diagnosed with a condition that upon further reflection, turned out to be benign.

Fast forward to today’s financial services industry and the term is now closely associated with the industry’s attempts to combat fraud. In this context a false positive is deemed as a suspicious transaction, loan application, or insurance claim, for example, which turns out to be innocent upon investigation.

"Taking the time to map the flow of data, and how any data matching or rule based decision system fits within an organisations processes, can often yield reductions in false positives that might have otherwise required much more sophisticated, and expensive methods to remedy."

When it comes to combating fraud one of the most challenging things to get right, regardless of which business sector you’re working in, is understanding how to minimise the amount of these false positives being generated.

In many respects, because of the huge growth in transaction volumes across all areas of financial services, precipitated by the advent of the Internet, most investigation teams place huge importance on reducing the amount of false positives their fraud defence systems are generating.

Many companies have complicated legacy systems to help them try and address fraud. Many just aren’t up to the job of coping with the exponential growth that has occurred in transaction volumes, and the volumes of data that they are being expected to deal with.

Additionally investigation capacity is a very finite, and specialist resource. Many companies deploying fraud detection systems become rapidly overwhelmed with investigations for these teams if they haven’t calibrated their solutions correctly – or configured their metrics in the most appropriate way for the business sector they are in. This results in a multitude of false positives that all require some kind of assessment.

Obviously fraud assessment or investigation requires resource, and takes time, which all adds significant cost to the business – and risks creating a drag on competitiveness while potential new business is held up in the approvals process.

This fact doesn’t help banks, insurers or financial providers who are under increasing pressure to increase profitability. Perpetually low interest rates, additional regulatory compliance burdens and increasing competition from new ‘fintech’ start-ups have all eroded their margins. As such these companies need solutions that improve their bottom line by reducing the cost of risks – not adding to costs or reducing competitiveness.

Unsurprisingly there isn’t a ‘one size fits all’ metric that a company can use as a bench mark to assess if its fraud defences are performing in line with a particular industry standard. Neither is there just one method that works as a cure all for all fraud teams to reduce their false positives.

While false positives typically range from anywhere from 2:1 to 25:1 in the finance and insurance sectors, according to statistics from the National SIRA database, the reality is that false positive rates are always going to be very specific to a sector, type of product, or even to a particular company. Each company will have a very unique or different level of risk appetite and profile of customer.

METHODS TO REDUCE FALSE POSITIVES FOR INSURANCE, BANKING OR FINANCE CLIENTS

There are a multitude of methods that companies can use to help them refine their fraud defences, to weed-out false positives, as well as create a more capable and effective fraud solution that will protect their business and their customers.

The key is having a solution in place that can weave these techniques into a holistic system, capable of integrating with legacy systems, to deliver a fraud defence that will reduce losses and make the investigation process as efficient and frictionless as possible.

ENHANCED RULE BASED MATCHING TECHNIQUES

Data matching and rule based decision systems are one of the most common methods of helping companies to identify and address fraud. Decision rules can be easily set up using a variety of parameters to help identify anomalies or group certain risk factors to help focus on applications that are more likely to pose a risk.

However, making rules based decision systems work most effectively, when it comes to reducing false positives, can be a challenge. If over simplified or deployed incorrectly rule sets can easily over-refer or flag inappropriate risks, and result in investigation teams being overwhelmed.

DATA ACCURACY AND ADEQUATE BUSINESS ANALYSIS IS KEY WHEN DEPLOYING A COMPANY'S FRAUD SOLUTION TO REDUCE FALSE POSITIVES

While increasing the sophistication of rule configuration filters by deploying velocity, matching, fuzzy logic, deterministic or probability based rules, the relevance and accuracy of data being fed into the system is subsequently increased.

This is a critical strategy, which organisations often overlook. All too often companies get excited about the possibilities that matching can provide to help them reduce false positives or address risk, only to discover that because of a lack of data validation in many of their fundamental systems, the quality or accuracy of their data severely restricts what they can achieve.

Another very important piece of the jigsaw to consider is the logical flow of data when determining decisions to flag risks.

Organsiations often have legacy or third party systems performing other checks. All too often they will forget to screen already identified, problematic applications from these systems, and then feed them back into their fraud systems. This then creates unnecessary false positives that can negatively affect the customer approvals journey.

It sounds obvious but many companies fail to take the time to perform adequate business analysis of the ‘eco-system’ within which their fraud solution is deployed. It is essential when deploying any additional fraud defence that a company makes sure that they are not creating issues that will reduce operational efficiency unnecessarily.

Taking the time to map the flow of data, and how any data matching or rule based decision system fits within an organisations processes, can often yield reductions in false positives that might have otherwise required much more sophisticated, and expensive methods to remedy.

INTELLIGENT COLLABORATION REDUCES FALSE POSITIVES AND REMOVES VULNERABILITY TO FRAUD

Another method that companies are increasingly using which proves highly effective in reducing false positives is to collaborate and share information on known risks. Accessing fraud intelligence databases that are populated by trusted industry partners is a highly successful mechanism that enables a company to refine their investigations processes, and score the likely level of risk an application poses much more accurately.

Enriching a company’s intelligence by adding a variety of third party, external, data sources to refine fraud referrals has proved tremendously effective at making sure that fraud defences are as effective as possible – and ensures that they are refined to reflect the idiosyncrasies that a particular sector or market segment represents.

Databases such as the National SIRA database provide opportunities for insurers, banks, telcommunications companies and retailers to gain access to a much wider view of known fraud risks that they couldn’t hope to have visibility of on their own. By adding intelligence from trusted partners into their application vetting processes, not only can companies significantly enhance their ability to reduce false positives but they also reduce their vulnerability to being targeted – as a result of the umbrella of protection afforded to them by the sharing of intelligence on fraud risk.

The National SIRA database has helped to prevent over £3 billion of fraud, which demonstrates how effective collaboration can be in the fight against fraud

Being part of a syndicated data service in this way can also help to give customers confidence that a company is doing all it can to protect itself from fraud, and so can provide additional brand trust in sectors that are often struggling to gain the respect of the markets they are seeking to serve when it comes to financial crime.