USPS Leverages Big Data To Fight Fraud
May 1st, 2013 By: John Rath
When you think “big data,” you probably don’t think of the United States Postal Service (USPS). As it processes more than 528 million pieces of mail each day, the USPS has become an active participant in the big data revolution, and operates one of the most powerful non-classified supercomputing databases in the world.
Residing in the Eagan IT and Accounting Service Center in Eagan, Minnesota is the core of the USPS supercomputing operations. The supercomputing and revenue protection program helps to cut fraud on its reported $65 billion in annual revenue. To detect fraud on 528 million pieces of mail each day the USPS started building IT architecture in 2006, and has been expanding it ever since. Simultaneously processing data from 6,100 mail pieces per second requires some hefty compute power.
“All of this happens in an average of 50 to 100 milliseconds,” says Scot Atkins, USPS program manager. “If something is off – a package with insufficient, duplicated or fraudulent postate – it is detected and intercepted in near real-time. A distribution clerk can reassess a package, errors can be tracked down, and fraud attempts – large or small – are reported to the U.S. Postal Inspection Service for further investigation.”
The system used to process and detect fraud is impressive. In a recent interview the USPS in-memory database was noted at 16 terabytes, up 6 terabytes from a 2010 Oracle case study report. Coupled with a transactional database it will perform billions of mail piece scans in a standard 15 hour processing day. Real-time scanning performs complex algorithms, powered by a SGI Altix 4700 system. Oracle Data Warehouse keeps fraud results stored in 1.6 terabytes of TimesTen cache, in order to compare and then push back into the Oracle warehouse for long term storage and analysis. Data ingest rates exceed 1.5 million records ingested per second per 16 core blade.
Fighting fraud, waste and abuse is top of mind for USPS and government organizations, as budgets are cut, and billions of dollars are lost in duplicative payments, overpayments and fictitious vendors.
sam 1Posted May 2nd, 2013
so what kind of fraud do they find?
JeffPosted May 2nd, 2013
“so what kind of fraud do they find?”
All of it. Or at least, that is the intended message the USPS wants to send. They describe it this way: “The biggest effect we’ve seen is deterrence. We can measure deterrence within our organization, and a large number of fraud cases dropped off significantly since we introduced a lot of revenue protection capabilities.” As always, the most money is saved in a category that’s almost impossible to measure: the fraud that wasn’t attempted in the first place.
ScottPosted May 4th, 2013
They ought to look at their employees to see what they’re doing. I’ve heard my carrier talking personal business on bluetooth. I suspect they are on their entire route.
RobertPosted May 8th, 2013
Actually this is a two-way street. If this process is as sophisticated as indicated, then it should be used to identify mailers who overpay postage. How many brides out there, who suspect their invitations may be over one ounce, automatically place an additional first class stamp “just to be sure it doesn’t come back”. Of course more knowledgeable mailers know that the second ounce is 20 cents. Consequently, over-an-ounce letters that should be going out at 66 cents, are consistently going out at 46 plus 46, or 92 cents. Accordingly, the average bride is overpaying 26cents per letter, and if the average bride is sending out 200 invitations, she is overpaying by 52 dollars. I am quite aware that the USPS has a 66 cent stamp in its offerings…even one with a wedding cake, probably to address this very problem. If this program and transactional database can detect an instance of underpayment in “milliseconds”, then why cannot it also be used to protect the overpayment by unsuspecting mailers.