Software AG Case Studies Faster Fraud Detection Drives Higher Profits

Edit This Case Study Record

	Faster Fraud Detection Drives Higher Profits Software AG

Faster Fraud Detection Drives Higher Profits

Software AG

Technology Category	Analytics & Modeling - Real Time Analytics
Applicable Industries	Finance & Insurance
Applicable Functions	Business Operation
Use Cases	Fraud Detection
Services	Data Science Services
Challenge	The company, a global clearinghouse for online and credit card payments, was facing significant losses due to fraudulent transactions. The existing fraud detection system could only handle 50 rules, but the risk management team wanted to add thousands more to enhance the accuracy of fraud detection. Additionally, the company wanted to reduce transaction time to achieve real-time fraud detection in all cases. The existing platform had an end-to-end Service Level Agreement (SLA) of 800 milliseconds, which was not met 10% of the time, resulting in either approval of the transaction or payment of a fine, costing an estimated $10 million each year. Read More
About Customer	The customer is a leading global online payments company that serves as a clearinghouse for online and credit card payments. The company operates on a global scale and is heavily reliant on its ability to effectively manage risk to maintain profitability. Fraudulent transactions were costing the company around 30 cents of every $100 in transactions. The company's risk management team identified that enhancing the company’s fraud detection algorithm with one additional rule would save $12 million annually. However, the company's existing platform could only handle 50 rules, and the team wanted to add thousands more to improve the accuracy of fraud detection. Read More
Solution	The company decided to implement BigMemory Max to improve its fraud detection capabilities. The existing platform, Oracle Exadata, was used to store fraud detection rules in disparate disk-bound databases, but it was limited in throughput and had inconsistent latency. A proof-of-concept implementation demonstrated that BigMemory could keep 100x the number of rules in a central, TB-scale in-memory store and boost throughput by over 30x, to 65,000 transactions per second. After confirming that BigMemory delivered extremely low, predictable latency at scale, the company was confident that BigMemory was the ideal choice for adding algorithmic complexity and accuracy, and also for meeting the new, stricter SLA. Read More Log in to view content
Contents

Technology Category

Analytics & Modeling - Real Time Analytics

Applicable Industries

Finance & Insurance

Applicable Functions

Business Operation

Use Cases

Fraud Detection

Services

Data Science Services

Challenge

The company, a global clearinghouse for online and credit card payments, was facing significant losses due to fraudulent transactions. The existing fraud detection system could only handle 50 rules, but the risk management team wanted to add thousands more to enhance the accuracy of fraud detection. Additionally, the company wanted to reduce transaction time to achieve real-time fraud detection in all cases. The existing platform had an end-to-end Service Level Agreement (SLA) of 800 milliseconds, which was not met 10% of the time, resulting in either approval of the transaction or payment of a fine, costing an estimated $10 million each year.

About Customer

The customer is a leading global online payments company that serves as a clearinghouse for online and credit card payments. The company operates on a global scale and is heavily reliant on its ability to effectively manage risk to maintain profitability. Fraudulent transactions were costing the company around 30 cents of every $100 in transactions. The company's risk management team identified that enhancing the company’s fraud detection algorithm with one additional rule would save $12 million annually. However, the company's existing platform could only handle 50 rules, and the team wanted to add thousands more to improve the accuracy of fraud detection.

Solution

The company decided to implement BigMemory Max to improve its fraud detection capabilities. The existing platform, Oracle Exadata, was used to store fraud detection rules in disparate disk-bound databases, but it was limited in throughput and had inconsistent latency. A proof-of-concept implementation demonstrated that BigMemory could keep 100x the number of rules in a central, TB-scale in-memory store and boost throughput by over 30x, to 65,000 transactions per second. After confirming that BigMemory delivered extremely low, predictable latency at scale, the company was confident that BigMemory was the ideal choice for adding algorithmic complexity and accuracy, and also for meeting the new, stricter SLA.

Impact #1	The company’s BigMemory-powered fraud detection platform now incorporates thousands of validation rules, transaction rules and risk rules, all stored in a 4TB BigMemory store.
Impact #2	The company now meets its stricter 650-millisecond end-to-end SLA for 99 percent of transactions.
Impact #3	The payment processor plans to expand its BigMemory store to 150 TB while committing to an even tighter end-to-end SLA of 250 milliseconds, which will deliver even greater results to the business.

Impact #1

The company’s BigMemory-powered fraud detection platform now incorporates thousands of validation rules, transaction rules and risk rules, all stored in a 4TB BigMemory store.

Impact #2

The company now meets its stricter 650-millisecond end-to-end SLA for 99 percent of transactions.

Impact #3

The payment processor plans to expand its BigMemory store to 150 TB while committing to an even tighter end-to-end SLA of 250 milliseconds, which will deliver even greater results to the business.

Benefit #1	Estimated savings of tens of millions of dollars in reduced costs from missed SLAs and fraudulent charges.
Benefit #2	Savings of an estimated $1 million annually in reduced database licenses.
Benefit #3	Boosted throughput by over 30X.

Benefit #1

Estimated savings of tens of millions of dollars in reduced costs from missed SLAs and fraudulent charges.

Benefit #2

Savings of an estimated $1 million annually in reduced database licenses.

Benefit #3

Boosted throughput by over 30X.

Download PDF Version

Overview

Faster Fraud Detection Drives Higher Profits

Operational Impact

Quantitative Benefit