As one of the earliest and most successful players in electronic payments processing, PayPal has become a household name that operates in more than 200 countries worldwide and processed over $450 billion in transaction volume in 2017.1 While overall, continued advances in technology are worth celebrating, they have also resulted in the increased sophistication of fraud. For PayPal, fraudulent activity manifests itself in the form of identity fraud, collusion, stolen accounts, and more.2 Modern day hacker attacks have increased in frequency and complexity, a trend that calls for counteractive solutions to thwart these criminal attempts. In response to this alarming phenomenon, the usage and integration of machine learning as a critical part of PayPal’s management of process improvement and product development is becoming increasingly important to the company’s continued success. Machine learning allows PayPal to differentiate itself as a secure platform with a fraud rate of only 0.32% of revenue, a mere fraction of the 1.32% average fraud loss at U.S. merchants.3 PayPal continues to make progress in detecting and preventing fraudulent transactions through increased investment in data science and machine learning research, resulting in a significant competitive advantage for the company as it seeks to differentiate itself and retain market leadership in an increasingly crowded sector.
In the short term, PayPal is building out its machine learning capabilities through significant internal development, as well as acquisitions of other external players. Internally, PayPal uses linear, neural network, and deep learning algorithms in combination to create feedback loops that work together to achieve higher accuracy of fraud detection. On the corporate finance front, in June 2018, PayPal announced its $120 million acquisition of Simility, a fraud detection and risk management startup that uses advanced machine learning to prevent fraud by identifying strange behavior that deviates from the norm.4 The scale effects of PayPal’s vast data sets manifests itself in the form of superior machine learning algorithms that improve in precision and reliability with every single transaction facilitated by the platform. This has allowed PayPal to become so advanced in fraud prevention to the point of being able to introduce PayPal One Touch, a platform that allows customers to check out from a merchant’s website without having to enter login credentials, a concept that seems like a security nightmare, but is simplifying the online shopping check-out process for countless users.5 This revolutionary platform was only made possible because of PayPal’s ability to combine the vast amount of consistent data it has on users’ regular transactions to create the ultimate security blanket.
In the medium term, PayPal is investing in further machine learning research that is focused on broader risk management areas such as improving credit decisions as well as attempting to minimize customer churn to improve revenue. From a regulatory perspective, using machine learning for the purposes of evaluating credit profiles is a lot more complex than fraud prevention, which is a challenge that the company is currently trying to work around.6 The PayPal data science team has been looking into developing predictive churn models using machine learning algorithms that analyze historical customer retention data to find potential patterns in churn-related activity.7 The insights gleaned can aid Paypal in identifying the product features that are critical to retaining customers and decreasing churn.
In addition to the existing ways that PayPal is using machine learning to improve its products and processes, I would recommend that they also tap into the predictive modeling potential of its vast amounts of customer transaction data. By building customer profiles through information on their purchasing habits, PayPal has the potential to greatly improve its marketing campaigns. Purchasing data is extremely valuable in terms of determining user habits and preferences. Armed with the knowledge on the unique ways that different customer segments respond to various forms of advertising, PayPal can develop more targeted and efficient marketing techniques that will hopefully result in increased revenue and user count, leading to even more valuable data that can be leveraged. The cycle can be repeated continuously, as more customer data will lead to more accurate pattern detection and predictive capabilities on all fronts.
In the context of PayPal, two important open questions related to this issue that I would welcome input on are:
- How should PayPal think about the various potential challenges around using machine learning in the credit profiling and evaluation process (both regulatory and non-regulatory)?
- Would it be more effective for PayPal to focus on further developing its machine learning capabilities in-house utilizing its existing technology and data as opposed to acquiring external firms that must be integrated into the ecosystem?
(Word Count: 763)
 Mikhail Kourjanski, “ML Data Pipelines for Real-Time Fraud Prevention,” QCON New York, June 2018. https://www.infoq.com/presentations/paypal-ml-fraud-prevention-2018. Accessed November 2018.
 “Driving Away Fraudsters at PayPal: Case Study,” H2o.ai, June 2018. https://www.h2o.ai/wp-content/uploads/2018/06/H20-Customer-Case-Study-PayPal-5302018-FINAL-NEW-2-1.pdf. Accessed November 2018.
 Michael Morisy , “How PayPal Boosts Security with Artificial Intelligence,” MIT Technology Review, January 25, 2016. https://www.technologyreview.com/s/545631/how-paypal-boosts-security-with-artificial-intelligence/. Accessed November 2018.
 Paul Sawers, “PayPal to Acquire Machine Learning-Powered Fraud Detection Start-up Simility for $120 Million,” Venture Beat, June 21, 2018. https://venturebeat.com/2018/06/21/paypal-to-acquire-machine-learning-powered-fraud-detection-startup-simility/. Accessed November 2018.
 Rik Kirkland, “AI is open for business: An interview with PayPal COO Bill Ready,” McKinsey Publishing, April 2018. https://www.mckinsey.com/featured-insights/artificial-intelligence/ai-is-open-for-business-an-interview-with-paypal-coo-bill-ready. Accessed November 2018.
 Eric Knorr, “How PayPal Beats the Bad Guys with Machine Learning,” InfoWorld, April 13, 2015. https://www.infoworld.com/article/2907877/machine-learning/how-paypal-reduces-fraud-with-machine-learning.html. Accessed November 2018.
 “Solving Customer Churn with Machine Learning: Case Study,” H2o.ai, July 2017. https://www.h2o.ai/wp-content/uploads/2017/03/Case-Studies_PayPal.pdf. Accessed November 2018.