To the texting generation even email is too slow, so the post office with its “snail mail” undoubtedly conjures up images of some decades-old, first-generation routing system. If you’re one who thinks the United States Postal Service (USPS) is behind the times, then it will come as a surprise to learn that this Ben Franklin-era innovation is using some of the most advanced computer technology of our time to analyze over 528 million mail pieces per day for signs of fraud.
USPS program manager Scot Atkins told Federal Computer Week that the post office has been using big data technologies since 2006 for their role in fighting fraud detection. Since that time, the agency has expanded its IT architecture to support “one of the world’s most voluminous real-time transactional data problems.”
The main USPS supercomputing facility, located at the Eagan IT and Accounting Service Center in Eagan, Minn., processes data from 6,100 mail pieces per second. The heart of the facility is a 16 terabyte in-memory computing system with a transactional database that can run comparative analysis on 400 billion records in the blink of an eye.
As pieces of mail from all over the country are scanned into the system, relevant data, including carrier and routing information as well as the item’s weight and size, are sent via the Postal Routed Network (a secure section of the Internet) to the database in Eagan. Each parcel’s dataset is compared with 400 billion records, while complex algorithms run through a validation checklist. Once the package data passes inspection, the information is sent back through the Postal Routed Network to the delivery center.
The entire process only takes “50 to 100 milliseconds,” according to Atkins. If the system detects “insufficient, duplicated or fraudulent postage” the package is intercepted immediately. All suspicious findings get passed to the US Postal Inspection Service for further investigation.
The USPS does not publicly disclose how much money is saved or reclaimed as a result of its real-time fraud detection system; however, given a yearly revenue base of $65 billion per year, even a small percentage would equal millions of dollars saved. Even more important is the message sent to would-be fraudsters. “The biggest effect we’ve seen is deterrence,” said Atkins, “A large number of fraud cases dropped off significantly since we introduced a lot of revenue protection capabilities as far back as 2006.”