The geek’s guide to 5 tech features that improve fraud prevention
Machine learning, artificial intelligence, a robust data network, customizable policies, alert management, and holistic customer journey protection are key components in the best fraud prevention software solutions.
These components vary by product and are essential for effective fraud prevention. But they aren’t the only things you should focus on. You should also assess the technical aspects of any potential solution.
“Most people don’t realize that a fraud solution can be a multi-year investment,” said engineer and Kount expert Conrad Kennington. “You have to train someone to use it, integrate with the software, and monitor and adjust it over time. It’s incredibly important to know the technical aspects of a solution upfront, but people often overlook them.”
A long-term fraud solution has many benefits, including increased profitability and growth opportunities. So to help you maximize your long-term investment and find the best solution, Kennington shared five technical fraud features and how they improve fraud prevention performance.
1. Uptime ensures system reliability
Uptime is the amount of time a system is operational. In other words, uptime measures how reliable a system is by tracking how long it’s down cumulatively per year in minutes, hours, and days. This system, as Kennington explained, is expressed in “number of nines.”
For instance, a system with an uptime of 90% or “one nine” has a downtime of 36.53 days per year or 2.4 hours a day. And a system that has an uptime of 99.9999999% or “nine nines” has a total downtime of 31.56 milliseconds per year.
A system with an uptime of nine nines is nearly impossible to achieve. All systems have at least some downtime due to maintenance or unforeseen issues throughout the year. In addition, a system’s hardware, storage ability, and infrastructure can affect uptime.
“Uptime speaks to a lot of things,” said Kennington. “How high quality is your code? How good is your infrastructure? How good are the engineers at keeping things up and running? Can the system handle fluctuations and spikes in high-volume traffic?”
Uptime is critical to consider in a fraud solution because downtime in service can mean opportunities for fraudsters and major losses. The higher the uptime, the more reliable the service.
As a benchmark, 99.95% uptime (three and a half nines) to 99.995% uptime (four and a half nines) is excellent system reliability. 99.999% (five nines) and higher is exceptional system reliability. You might reconsider your chosen solution if it has anything below three nines uptime.
2. Low latency improves customer experiences, retains sales
Latency is the delay in processing data over a network connection. In essence, it’s the real-time speed to decision a transaction. For example, in digital fraud prevention, latency refers to the round-trip time, measured in milliseconds, that it takes for a solution to approve or decline an order after a customer checks out.
This process includes the time it takes to calculate risk, relay to third-party connectors if needed, and send information to the merchant. The faster a solution is at decisioning, the lower the latency and the better experience your customer will have.
“Real-time response is kind of a buzzword in the industry, but latency makes a big difference when customers are actually interacting with your solution,” said Kennington.
A delayed or latent response can cause friction for customers and lead to cart abandonment, accidental double purchases, chargebacks, and lifetime loss of a shopper. For example, in 2006, Amazon revealed that 100 milliseconds in latency cost them 1% of sales — around $1.6 billion at the time.
Acceptable latency rates depend on the industry. For digital fraud prevention solutions, around 250 milliseconds for a real-time response is about as fast as it gets. But a real-time transactional risk response of 500 milliseconds or more can dissatisfy customers and lead to cart abandonment.
3. Device intelligence provides more accurate risk assessments
Accuracy refers to how precise a fraud model or solution is at detecting fraudulent activity. Overall, determining accuracy is difficult. An accurate solution will, at minimum, use artificial intelligence, supervised and unsupervised machine learning, and a robust identity trust data network. These things combined provide greater accuracy when preventing current and future fraud attacks.
Equally important for accuracy, though, is the type of data a solution collects. For example, payments data, location ID data, unique customer data like loyalty numbers, and digital identifier data like email addresses provide robust insights for assessing transactional risk.
However, these types of data are now table stakes for fraud solutions. So Kennington says you’ll want to look for a solution that also offers device intelligence capabilities or device fingerprinting.
Device intelligence and device fingerprinting combine attributes of a device like its operating system, screen resolution, language setting, web browser type, and incognito modes to create a unique identity. Solutions that collect device information can recognize its identity elements wherever they appear.
“When people commit fraud, they tend to do it over and over again from the same device,” said Kennington. “They may be able to change their phone number or email address, but the deeper you go with device collection, the harder it is to change things.”
Device fingerprinting or device data collection happens at levels. The deeper level that a solution collects data, the better. Though a company may not reveal its collection level, Kennington recommends solutions that collect beyond the superficial.
4. In-house data scientists continually improve fraud models
A data science team uses statistical methods and other tools to analyze data and create predictive models that stop emerging and future fraud. Data scientists constantly improve machine learning fraud prevention and device fingerprinting models.
“If you don’t have a data science team that’s dedicated to continually improving models, fraudsters are going to figure it out, and they’re absolutely going to get around your solution,” said Kennington. “These models keep solutions at the cutting edge and help them predict attacks.”
For instance, even though people tend to commit fraud from the same device, they also change devices frequently. Kennington says this is one reason why device data collection is getting more difficult, and anonymity is becoming easier on the internet. Data scientists need to continually improve fraud models to stay in front of these issues.
Kennington recommends finding a solution with in-house data scientists dedicated to machine learning and device data collection. Ideally, he says, a solution would have teams dedicated to each exclusively.
5. Scalable solutions accommodate business growth
Scalability allows businesses to grow at size and scale or take on more customers faster and more efficiently. The ability to scale is essential for an effective solution.
Old code, limited data centers, and weak system architecture can hold a business back as it grows. And a solution that isn’t scalable can expose it to serious risks.
“A scalable solution grows as the business grows,” said Kennington. “A lot of companies are behind in this area right now, and people don’t realize it until they start expanding their business. You want a solution that adapts to you, your traffic, buying seasons, and purchasing behavior.”
Retail e-commerce sales reached USD$4.9 trillion worldwide last year. And this figure is set to grow by 50% by 2025. Merchants have exciting times ahead of them, but they need to make sure they’re ready.
Kennington recommends businesses inquire as to whether a solution is organized as a microservice architecture application so that it can scale horizontally. Horizontal scaling is similar to opening up more lanes on a freeway as traffic volume increases. When your system scales horizontally, it can expand to prevent bottlenecks and accommodate growth.
Outdoor retailer achieves global expansion with the right fraud solution
With innovative designs and uncompromising quality, the outdoor retailer, Arc’teryx, is the outfitter of choice for many adventurers. But as it expanded into e-commerce sales, rising fraud and the threat of chargeback monitoring programs slowed growth.
So Arc’teryx upgraded its fraud solution, selecting Kount for its global network data, customer service, and User Experience Engine. Overall, the company achieved strategic geographical expansion, reduced its chargeback rate to .35%, and reached a historically low manual review rate – all while crafting personalized shopping experiences for its customers.
“Kount’s been a huge help to our fraud prevention, and it’s been evolving with us as well,” said a fraud prevention specialist at Arc’teryx.
Kount satisfies a number of fraud use cases with advanced AI equipped with unsupervised and supervised machine learning to deliver accurate decisions in less than 250 milliseconds. And they analyze billions of data points from the Identity Trust Global NetworkTM to reduce chargebacks and help businesses capture more revenue.