3 ways to customize AI to achieve your business goals
More businesses are using artificial intelligence (AI) and machine learning to stop fraud. And it’s not surprising why. AI can do what humans can’t: identify fraud at scale and in real time. Plus, AI that uses supervised and unsupervised machine learning adapts on its own to catch new fraud. But AI isn’t the only tool that businesses need. Enterprise businesses have complex operations. And they need more than one strategy to fine-tune fraud prevention and achieve specific business outcomes.
“It’s really about cost and performance optimization,” said Brady Harrison, Kount Senior Data Analyst, at Kount’s Fall 2020 Digital Protection Summit.
For businesses that work in multiple channels and locations, “optimal” fraud prevention depends on business goals. A fraud prevention solution that makes unclear decisions limits businesses. Depending on their product or service, businesses may have different tolerance levels for fraud, chargebacks, manual reviews, and false positives. To maximize their business, these companies need maximum control over their fraud prevention. That’s where AI, combined with a flexible policy engine, can help.
“GNC Canada might not be the same as GNC.com, and we understand that,” explained Rowdy Durci, GNC Senior Loss Prevention Program Analyst, at the Digital Protection Summit. “We utilize that information to benefit us in the best way.”
Businesses take advantage of policies to focus their fraud prevention on specific, unique results. Advanced AI can improve those policies in two ways: identifying policies that work and ensuring that they deliver the best outcomes.
1. Reduce risk with AI and machine learning
Effective fraud prevention starts with AI and machine learning. In a new sales channel or region, risk is a balancing act. Too much authentication can frustrate customers, but too little can lead to a rush of chargebacks. And chargebacks are just one in a range of possible problems.
Machine learning can help balance risk in four connected areas: inventory, new channels, chargeback volumes, and digital accounts. For example, businesses that place a high priority on launching new channels can customize chargeback tolerance in that channel.
AI fraud prevention that combines both forms of machine learning uses historical information (“supervised” machine learning) to stop known threats and reduce risk. AI can also catch new attacks (“unsupervised” machine learning) by scanning for unusual activities. In this way, businesses can worry less about fraud as they grow.
“Nobody wants a new channel to go live that has so much friction that adoption is poor, or they just get killed in chargebacks,” said Harrison. “So it’s not a trial by fire every time we launch a new product or channel, we have some historical experience based on how our customers perform, and then we can leverage that.”
2. Target specific results with policies based on AI
AI, combined with an identity trust network, can accurately assess the risk of each interaction. But sometimes risk depends on context. For example, a large, recurring order might go to a known distributor. But it could also stock a gray market reseller. As a result of the coronavirus pandemic, GNC saw customers suddenly change their behavior. Purchases of vitamin C, multivitamins, and zinc spiked.
“It was important to understand that those items weren’t necessarily for fraud,” Durci said. “But we also needed to control the inventory because customers would buy in large quantities to try to resell later.”
GNC needed to solve contextual problems to manage its supply chain. But any solution had to be flexible and adjust to nuances at GNC. For example, purchases from China required a unique approach, while resellers posed special challenges that varied by location.
Policies help solve these problems. Machine learning does the core risk analysis, and policies relate that risk to the individual business and its goals.
3. Adapt trust to the interaction, need, and goal
What does successful fraud prevention look like? It depends. Some businesses need to emphasize inventory control. Others need to reduce chargebacks or manual reviews. To achieve the best results in a complex business model, policies can refine decisions and improve automation.
Whatever the goal, fraud can take different forms between companies and channels. Machine learning helps identify unique trends related to locations, accounts, devices, and more. Policies allow businesses to apply those trends in nuanced ways to different parts of their operations.
By adjusting trust levels to the channel or region, businesses can quickly reduce manual reviews and chargebacks. They can also stop card testing attacks, balance inventory, and reduce customer friction.
Kount’s adaptive AI helped GNC reduce chargebacks and manual reviews
Enterprise companies often approach fraud prevention with specific ends in mind. For example, GNC wanted to manage orders shipping to China, control purchase volumes, add new payment card types, and protect a new BOPIS channel.
Unsupervised machine learning helped them identify likely fraud immediately. Kount’s unsupervised machine learning links billions of interactions to determine risk in real time.
Supervised machine learning allowed GNC to refine their protection. This type of machine learning was trained on both GNC and Kount datasets, so it tuned the models to improve accuracy.
Policies based on machine learning optimized results. Each policy was specific to trends in GNC’s channels. Those policies helped improve Kount’s transaction safety rating, Omniscore, for every GNC transaction.
This layered approach allowed GNC to beat industry benchmarks in every area, including a steep reduction in chargebacks and manual reviews.
Achieve your business goals with Kount’s AI and machine learning
Kount’s AI combines supervised and unsupervised machine learning models to make accurate decisions at speed and scale. Those models are powered by the Identity Trust Global NetworkTM. Kount’s network links 32 billion annual interactions across 250 countries and territories, over 75 industries, and over 50 payment processors and card networks.
Plus, Kount’s decision engine allows businesses to create policies in minutes to address unique business needs and emerging attack methods.
Together, Kount’s Identity Trust Platform and decision engine help businesses quickly reduce chargebacks, false positives, and manual reviews. And it helps them target specific goals and outcomes for superior results.