Last Updated:

What to consider before buying / designing a fraud detection system?

Fraud detection

As payments are increasingly executed using mobile devices, the infrastructure is changing. As always, a multitude of banking channels, financial services providers, payment processors, and payment networks are jockeying for position in a highly competitive ecosystem. Whether you are designing a fraud detection system or evaluating currently available systems in the market, read this article to discover the important aspects before your big decision.

Network participants must adapt to an increasingly wide variety of payment mechanisms, devices, and types. Not long ago, nearly all payments originated from a relatively small number of point-of-sale terminals. Now payment processors must efficiently handle requests from billions of mobile devices worldwide. These transactions must traverse a complex payment processing network at near real-time speeds.

Architecture for Modern Payment Processing

Legacy payment processing infrastructure is not optimized for the new tsunami of  transactions. Facing a massive increase in mobile payments and consumer demand for real-time processing, these systems are struggling to keep up with time-sensitive, high-volume data.

Slow processing speeds can frustrate customers and merchants, encouraging them to change payment methods or even vendors. Poor performance also increases the likelihood of potentially costly and catastrophic data processing failures. Large companies that process high volumes of data are especially sensitive to a server or system failure. Downtime can negatively affect user satisfaction and may result in fines, potential regulatory oversight, and a loss of business.

People expect to conduct transactions wherever they are, at any time, using their  mobile devices. However, some payment processing systems lack adequate edge processing capabilities to meet these needs. Scalability can also be a problem, with legacy infrastructures unable to scale to handle traffic bursts. For example, seasonal traffic and peak loads—including spikes that occur during events such as Cyber Monday can overload some systems.

In the payment hub, compute speeds must be exceptionally fast. The solution should include a high performance operational data store for fast lookups, allowing the system to quickly and accurately assess transaction validity. This store cannot be based on disk-based data platforms due to the overhead that negatively impacts latency.

Focus on Payment Hub Services

Preventing Fraud Is a Balancing Act

Not only do consumers expect their purchase transactions to be completed without hassles or delays, but they also trust payment processors to protect their accounts, data, and assets from fraud. Yet the increasing volume of online and mobile transactions and payment methods drives up the number of fraud attempts.

Businesses want to protect themselves from fraud, but they must strike the right balance between speed and accuracy. When making a purchase, consumers expect the transaction to be processed in two to three seconds. However, network transit time consumes much of this window, requiring payment processors to complete their part of the transaction within milliseconds.

Accuracy is the other critical objective when processors run checks for verification, authentication, and fraud. In the effort to return results quickly, processors might use simple fraud algorithms, but they frequently return false positives where a transaction is incorrectly flagged as fraudulent. When cardholders are delayed by the return of a false positive, they often switch to a different payment method, such as another card. The merchant still gets the sale, but when the processor’s fraud detection algorithm is too strict, the card-issuing bank often loses the transaction fee.

As a result, today’s payment processors need to turn to more sophisticated fraud algorithms. In many cases, they must rely on multiple fraud algorithms to assess several
scores to most accurately predict fraud. This extra processing load requires enhanced performance to stay within service-level agreements and avoid overloading the system.

Even legacy payment processing solutions that work satisfactorily will soon be  overwhelmed by burgeoning workloads and the resulting increases in latency. What’s more, these last-generation solutions offer no competitive advantage or opportunities for innovation. That’s unacceptable in a market that clearly rewards first-movers.

To handle increased volumes and workloads, companies need to modernize payment processing systems. You need solutions that can support the business and meet the industry’s changing technology and consumer requirements—without imposing  unnecessary delays processing backlogs.

Source: Javelin Research, “False-Positive Card Declines Push Consumers to Abandon Issuers and Merchants,” August 27, 2015

Choose Technology with Essential Functionality

A real-time payment processing and fraud detection solution should include the following capabilities and features.

  • Processing speeds at extreme scale. An operational, in-memory computing platform that manages data using in-memory storage is best to deliver microsecond-level speeds at scale. It also performs parallel execution for fast application speed and supports ultra-low latency.
  • Real-time processing. Processing should be integrated with an in-memory data store. This feature helps you process data at the moment it is generated or upon ingestion. It also offers real-time classification and prediction workloads at scale, which is especially important for live transactions.
  • Modern fraud detection methods. Solutions should be able to run multiple fraud detection algorithms and use the results to make the most accurate decisions rapidly. High-performance solutions can run multiple algorithms in less than 30 milliseconds, which is one-tenth of the time it takes to blink.
  • Elastic, seamless scalability. A solution that can scale up and down without interrupting
    jobs helps you keep pace with changing workloads. You should be able to upgrade it without delaying task processing to ensure continuous operation.
  • Low-latency batch and stream processing. An application-embeddable data processing
    engine ensures that all tasks are lightning fast. Since the collection of payment transactions is essentially a stream, a powerful stream processing engine can handle the transactions as they are delivered into the system. A complementary batch processing system lets youprocess historical data to leverage it as part of the payment verification and fraud analysis tasks.
  • Stability. To be reliable, solutions must always be available for processing. In-memory data replication provides a robust yet performant means of fault tolerance. In the case of a temporary outage of the entire cluster, a hot restart feature helps you get all nodes up and running again quickly without having to reread all data from the original sources.
  • Enterprise-grade security. As mobile payments introduce many more access points to the processing network, the need for security rises. Authentication and role-based access controls protect data from unauthorized viewers, while encrypted data transmission helps ensure data privacy.
  • Intuitive manageability. By providing tools to monitor your system, a management console can help you ensure service-level agreements are met. It also provides the tools needed to troubleshoot problems when they arise, reducing the cost of downtime.

In-the-Moment Processing Accelerates Time to Results

A modern, real-time payment processing solution offers valuable benefits across your organization. From a business perspective, the right solution helps you deliver faster responses and improves time to results. Modern technology can ingest, categorize, and process vast amounts of data with ultra-low latency, helping you support continuous intelligence practices. It can also help you process time-sensitive data from numerous sources, where and when it is generated. With this unprecedented speed, you can make faster, smarter business decisions.

The ability to process increased volumes of high-frequency data also gives you the opportunity to use artificial intelligence and machine learning to enhance business insights. You can feed huge data volumes to powerful machine learning models, yielding valuable new correlations and patterns that can be beneficial to your business.

A modern payments processing solution also simplifies technology deployment and operation. A single, lightweight system can elegantly address even the most challenging architectural requirements. By choosing an integration-friendly solution, you can avoid adding further complexity or massive administrative overhead to your environment.

A solution designed for simplicity reduces the cost and complexity associated with multi-component and multi-system architectures. It also lowers your need for multiple skill  sets, simplifying staffing.