Case Studies How One Global Digital Payment Provider Drove Growth by Improving Fraud Detection, Reducing False Negatives and Strengthening Customer Satisfaction
Edit This Case Study Record

How One Global Digital Payment Provider Drove Growth by Improving Fraud Detection, Reducing False Negatives and Strengthening Customer Satisfaction

Analytics & Modeling - Predictive Analytics
Application Infrastructure & Middleware - Database Management & Storage
Infrastructure as a Service (IaaS) - Cloud Storage Services
Finance & Insurance
Business Operation
Fraud Detection
Software Design & Engineering Services
System Integration
Creating a more accurate fraud detection and prevention solution required factoring in unique attributes of the payment provider’s business model. The solution had to consider a myriad of realities, which included the company’s payments, rewards, and purchase protection programs, all of which have different rules and data-processing requirements. Unique profile attributes for 150 million customers, including payment and invoicing preferences, user ID, IP address, devices, location data etc. Mobility and multiple device dynamics that are often entry points for criminals to target mobile and “cross border” transactions. Visibility into payment histories and behavior patterns that help determine the validity of a transaction. Legacy database and infrastructure strategies that are unable to handle the requirements of a contemporary application, including leveraging near real-time data as a critical component of modern fraud algorithms.
Read More
The customer is a leading global payment provider that has made fraud detection, intervention, and prevention a top priority. With the rise in digital payment options, the company faces significant challenges from cyber criminals. The cost of online fraud has been growing, and it not only affects payment providers but also merchants, as victims of fraud often avoid certain merchants afterward. The payment provider has a complex business model that includes payments, rewards, and purchase protection programs, each with different rules and data-processing requirements. The company serves 150 million customers with unique profile attributes, including payment and invoicing preferences, user ID, IP address, devices, and location data. The company also deals with mobility and multiple device dynamics, which are often entry points for criminals targeting mobile and cross-border transactions.
Read More
The payment provider decided to leverage Aerospike, a high-performance NoSQL database, to build a more sophisticated and accurate fraud application. Aerospike is a key-value store built explicitly to run on SSDs, delivering speed at scale. The solution required defining a new set of database requirements to support an application designed to leverage massive data and real-time workloads to score fraud. Aerospike needed to meet a 750 ms SLA to evaluate fraud on every payment transaction, seamlessly and cost-effectively scale up to handle consistent growth in data and objects, improve query performance and data consistency, maintain consistent high availability in a 24x7 digital payments environment, avoid a costly rip-and-replace scenario, and greatly reduce data load times while providing reliable access to fresh data. The solution was built on Aerospike’s high-performance NoSQL database developed explicitly for flash storage to deliver superior speed, horizontal scalability, reliability, and lower TCO.
Read More
Aerospike’s unparalleled combination of speed and scale enabled the payment provider to write a more comprehensive fraud application that leverages a significantly larger data set to power a more sophisticated rules algorithm.
The net result is improved profitability as a result of more accurate fraud identification and a reduction in customer churn.
Aerospike consistently met the 750 ms SLA for fraud prevention for every financial transaction.
Accelerated load speeds to > 1 million writes/second, with support for stream loading while reading from the same database.
Improved application read latency to ~250 microseconds.
Reduced total query time (at 200 DB queries per transaction) to ~75 milliseconds.
Download PDF Version
test test