October 24, 2017
Keeping fraud costs at bay: improving efficiency and saving costs with adaptive behavioural analytics
Featurespace Subject Matter Expert Roger Lester explores how advanced fraud management systems can help organisations make cost savings and minimise losses, while making significant improvements to their fraud teams’ efficiency.
Not many months go by without new headline fraud statistics showing the growing challenge that fraud poses to industries including financial services, insurance, and gaming.
However, these statistics do not always reveal much detail about what factors contribute to total fraud loss, above and beyond the fraud attacks themselves. In my experience working in the financial services industry, there are three main sources that fraud costs can be attributed to when analysing the financial and operational impact of ‘successful’ fraud attacks:
1. Loss of Goods/Services: the loss cost most often associated with fraud attacks. Ultimately, the loss of physical goods or services is the obvious expense that results from fraudulent activity.
2. Initial cost of the fraud attack: this is the cost immediately felt by a business, before fraud is identified. A credit or debit card is used by a fraudster, who disappears with the goods before the fraud is detected. Typically, the Card Issuer will refund the customer straightaway. The Card Issuer might then issue a chargeback to the merchant. In these instances, there is a risk to the Acquirer, because any losses which cannot be paid by the merchant will be the Acquirer’s liability.
3. Operational costs: fraud attacks need time to review and require investigation by staff members, which is costly to businesses. This includes the cost of investigating ‘false positives’ – where genuine customer activity is incorrectly flagged as fraudulent and inconveniences the customer.
The extent of the fraud challenge
Each of these cost types adds significantly to the total fraud costs organisations are facing. The scale of the challenge becomes visible by looking at some headline figures:
How does the rising fraud threat effect operational teams?
As a fraud team manager, I witnessed first-hand what impact the volume of alerts has on staffing costs and planning, especially if the alert volume fluctuates a lot. To tackle these issues, fraud teams have to deal with peak alert volumes.
This leaves an organisation with two options:
1. Staff your team to meet peak alert volumes. These resources will have to be reallocated when alerts are low, without causing too much disruption to projects. In reality, this is impractical and a challenge to maintain.
2. Adjust your team size according to the average number of alerts – and either allow for overtime, or adjust the risk level during peak times (e.g. only investigate everything over $1,000, instead of $500). However, this approach can result in overworked agents or alerts that do not get investigated efficiently, exposing your business to increased fraud loss risk. I know from personal experience that many organisations will not give fraud managers additional headcount unless there have been significant fraud losses. Therefore, option two above is the most common approach taken by businesses.
How can organisations improve their operational efficiency and minimise fraud losses?
A cutting-edge, machine learning fraud prevention system solves this dilemma by understanding the behaviour of every individual in real-time and spotting anomalies as they occur.
Using advanced machine learning to tackle fraud, a financial services organisation I have worked with:
- Reduced their alert volume from over 1,000 per day to under 200 per day– a staggering 80% reduction
- Improved their false positive ratio (the number of genuine accounts blocked incorrectly for every genuine fraud attack caught)
- Improved their fraud detection rate
- Achieved a more consistent volume of alerts, at a much more manageable level
- Reduced staff overtime managing alerts and absorb more business without the need to request more headcount
What should businesses do next?
To tackle fraud and simultaneously reduce the associated costs, organisations that want to get ahead are investing machine learning systems that use Adaptive Behavioural Analytics. These fraud systems – like Featurespace’s ARIC Fraud Hub – improve fraud detection, reduce genuine transactions declined and reduce associated fraud costs, working automatically 24/7, 365 days a year.
Typically, a return on investment of an advanced machine learning system can be experienced within months of deployment – especially where multiple, siloed fraud systems can be combined into one efficient, cost-effective fraud solution.
Advanced fraud systems are the way forward in giving your ‘good’ customers an outstanding experience, while stopping fraudsters in their tracks.
Roger has worked in the payments industry for more than 30 years and is currently a Featurespace Subject Matter Expert in Payments. Having worked both with Card Issuers and Acquirers, Roger brings his industry expertise and insight to ensuring that Featurespace’s ARIC Fraud Hub enhancements match the risk management needs and requirements of acquirers, merchants and payment processors in the financial services sector.