News, Education and Events Decoding Digital Payments & Fraud

News, Education and Events Decoding Digital Payments & Fraud

Machine Learning and Manual Review: Validate Identities and Verify Good Orders Faster

Machine Learning and Manual Review: Validate Identities and Verify Good Orders Faster

By Beth Shulkin, Senior Director of Global Marketing, Whitepages Pro

In recent years, the popularity and utilization of machine learning has grown exponentially. Many e-commerce and retail companies have started using machine learning to optimize their fraud workflow management and automate their processes. Even though machine learning is slowly becoming the standard as it further amasses computational power and improves the quality of its training datasets, merchants still need humans to help catch fraud in the manual review process.

The overarching goal of manual review is to balance accuracy and efficiency and, these days, manual review is becoming more and more optimized. The challenges for e-com businesses and review agents in the world of e-commerce, however, still remain. Review agents must help merchants prevent fraud, verify good orders faster, and increase revenue.

First, a big challenge for merchants is keeping up with increasing customer expectations of mobile commerce. In other words, preventing customer friction is now an even bigger essential as online shopping is the top activity performed on a mobile device for consumers— they expect instant results wherever, whenever. In today’s fast-paced tech-driven world, convenient shopping with the click of a button is the basic standard or consumers will switch to another merchant.

Second, merchants face increasingly sophisticated online fraud attacks. For example, credit card chargebacks are a costly problem for online merchants. Hefty fines from credit card companies should be a thing of the past; e-com businesses need to determine the good customers quickly and find potentially fraudulent orders that really need review.

Lastly, e-com businesses with a manual review process will not scale effectively due to how much the process costs and how long it takes. More orders as the digital side of business takes off means more time spent, more agents required, and more money spent on manual review.

There are so many tools that can be used during manual review, from regulatory and compliance-related tools to ones that help with workflow. Before spending money on expensive data queries, companies should research and learn more about manual review solutions, and at the basic level of manual review sits identity data.

Leveraging Identity Data and Machine Learning During Manual Review

To address these challenges head-on, more companies in the world of retail and e-commerce are striving to meet consumer expectations while keeping up with dynamic fraud strategies to minimize losses, boost revenue, and control operational costs. This means turning to manual review solutions that enable faster, often much faster, processing times per order.

Therefore, review agents need access to multiple internal and external tools that enable them to determine the legitimacy or risk of an identity behind an order before resorting to more expensive methods. Manual review agents must be armed with the proper tools to help them focus on the right insights to make a decision, or they may delay or reject legitimate customers, thereby causing negative customer experiences. A top-notch manual review tool should have:

  • Both digital and traditional identity data – A full view of the risk associated with the identity behind a transaction needs to provide clear, actionable insights to help a review agent make a decision, such as name-to-number match, phone carrier information, email domain validity, and associated people.
  • Machine learning insights – Machine learning trains, refines, and leverages networks of data and learns from these patterns of data to produce a risk score. It is impossible for a manual review agent to do the same as accurately and efficiently.
  • Positive and negative risk signals – A concise list of primary factors should take all transaction inputs into account and explain to an agent how they influence a risk score.
  • Full identity comparisons – These comparisons should show specific attributes for matches, mismatches, and invalid inputs based on the links between a transaction’s identity elements. The comparisons need to provide linkages, such as linking email to name, name to address, and more.

To address the challenges of fraud, more companies in the world of retail and e-commerce are striving to meet consumer expectations while keeping up with dynamic fraud strategies to minimize losses, boost revenue, and control operational costs. A streamlined, frictionless experience also improves the odds of repeat customer business, which is the easiest way to lower the cost of customer acquisition across the board. This all means using manual review solutions that enable faster, often much faster, processing times per order. As the speed and sophistication of online fraud increases, the fraud prevention process requires more repetitive tasks that must be completed faster, more accurately, and more efficiently than humans can manage; a comprehensive solution powered by machine learning and network data is the next best step.

E-commerce companies need to take advantage of machine learning tools to prevent fraud and produce nearly instant, highly accurate results. Humans’ intuition may never be replaceable, but if manual review agents can be supplied with insights garnered from a robust machine-learning operation, merchants can save time and costs while improving customer experience, accuracy of review, and dramatically cutting down on fraud attacks and chargebacks.

Beth Shulkin Whitepages ProBeth Shulkin is the Senior Director of Global Marketing at Whitepages Pro where she’s responsible for strategic oversight and leadership of product marketing, partner marketing, and sales enablement. She has over 15 years of experience in product and product marketing GTM, spanning several industries including financial, call center software, direct marketing platforms, and data API technologies.

CNP Fraud Accounts for 85% of  All Card Fraud in Australia