Argyle Data: Re-Writing the Book on Fraud Analytics

CIO VendorTom Ryan, President & CEO
Fraud has become both front page news and part of the board level agenda. The Association of Certified Fraud Examiners (ACFE) found that a typical organization loses 5% of revenues to fraud each year. The more connected the industry the more potential there is for fraud. The mobile industry alone loses over $40 billion annually to fraud. When it comes the detecting fraud in the financial services industry 16% take up to 8 hours to detect it and 22% do not know the avarage time for detection. Fraud is often described as a triple whammy: (1) You lose money on the fraud itself. (2) You lose money on the fines and (3) You lose money due to brand damage. Traditional approaches, designed to combat fraud, often bring together a patchwork quilt of database, ETL (Extract, Transform and Load) and BI technologies working in a batch mode. The problem is this delivers insights that are hours or days behind. Unfortunately fraudsters work in real time and often complete their job in under a day.

Coming to the aid of such companies is California based Argyle Data, which uses state-of-the-art machine learning on a Hadoop stack to deliver fraud analytics applications, that can ingest data and analyse it in real time, reducing the window for risk or fraud from hours (or days) to minutes. Under the leadership of Tom Ryan, President and CEO, Argyle Data has built a team of experts on on applying machine learning on Hadoop at massive scale complemented by industry fraud experts. Prior to Argyle, Tom was president and CEO of Alpine Data Labs, a leader in predictive analytics for big data.
Delivering Fraud Specific Solutions to Mobile Communications, Finance and eCommerce

Working with some of the biggest players in mobile communications, financial services and ecommerce, Argyle’s products, use technology pioneered at Facebook and the NSA to deliver machine learning at petabyte scale to identify fraud and anomalous beheviour. Tom Ryan uses the phrase “When minutes means millions” to drive the prioritization of the company’s fraud analytics strategy.

ArgyleDB is an integrated solution that brings together all of the critical components to drive fraud analytics from the 1990’s to the era of machine learning and big data. They identified packet ingestion, schema-less indexing with time-series support, anamoly detection with machine learning and querying at petabyte scale as key requirements. ArgyleDB Ingest is able to non-invasively tap into the network and perform Deep Packet Inspection (DPI) and stream live network packets, or log files into a “Data Lake”. The data is stored in a key value database deveoped at the NSA that indexes the data in a schema-less way with time series support. ArgyleDB Machine Learning is able to access streaming data and perform anomaly detection using multiple algorithms that learn online against the full data set (both streaming and historical). This is critical to identifyiing fraud patterns where comparisons to a previous year/season is critical. ArgyleDB Query is a native Hadoop real-time SQL database that supports schema on read and queries the data lake through ANSI SQL. Complex joins, aggregation and windowing functions, that are critical to identifing fraud, are supported across 1000’s of nodes.

Roadmap for Argyle Data

Looking towards the future, Argyle Data’s mission is to re-write the book on fraud analytics and create a data lake on which to build a suite of related applications. The company also intends to take their unique ability to counter fraud, to additional customers in U.S.A and Europe, and plans to expand in Asia as well.

Argyle Data

Tom Ryan, President & CEO

A provider of real-time analytics at network speed and Hadoop scale for data-driven organizations in telecom, finance, and e-commerce