If the Postal Service was a private company, it would rank 37th on the Forbes 500 list. The massive agency collects more than $70 billion in annual revenue and has more than 600,000 employees.
By contrast, USPS Office of the Inspector General that oversees USPS has about 1,000 employees and fewer than 400 contract workers.
However, a USPS OIG data analytics shop of about 50 employees has been using the agency’s huge troves of data to detect fraud, waste and abuse that might otherwise go undetected.
“How can you possibly oversee an entity that size in an effective way if you’re not using data analytics? You can’t,” Kelly Tshibaka, the USPS OIG’s chief data officer, said at a Data Foundation summit last week in Washington.
In the last year, the data analytics team has helped auditors avoid making more than $110 million in improper payments and recover $121 million in fines and restitution.
More recently, the USPS OIG has used data analytics to pick up anomalies in vehicle maintenance contracts. The Postal Service has a fleet of more than 470,000 mail delivery trucks.
“There’s this contractor who seemed to be spending a lot on the same vehicle,” Tshibaka said. “Why is that mail truck getting that same service performed over and over?”
Upon closer examination, the watchdog office found that contractors had been billing multiple times for maintenance work on a single truck — even though the work wasn’t actually performed. The OIG estimates the contractor received more than $61,000 in fraudulent payments.
“Now $61,000 in the bottom line of a $70 billion revenue agency is nothing, but when you think about 470,000 vehicles, and we’re looking for $61,000 across many line items, because that’s what our analytics can do, that adds up quickly,” Tshibaka said.
To that point, the USPS OIG launched three investigations into a contractor for custodial services that uncovered more than $152 million in fraudulent payments.
“It went over many years, because you have to look at all these little line-item transactions for cleaning service at a time — millions of transactions — and that can only be done through data analytics,” Tshibaka said.
“What we’ll sometimes hear people say is this is dirty data,” Tshibaka said. “Is there something wrong with it? No. Is it contaminated? No. It just takes time to put it in order so you can actually use it, and then you’ve got to put it in order so that this data will talk to your other data so that you can actually do analytics on it.”
However, about 80 percent of the data analytics process in the OIG requires sorting of this dirty data, some of it dating back more than 200 years, by database architects. Tshibaka compared this data janitorial (or “datatorial”) proccess untangling jewelry.
“You have no jewelry to work with until it gets untangled,” she said. “You have no data to work with until it gets untangled.”
However, the OIG has made progress by establishing a “data dictionary” for common data fields like first name, last name and employee identification number.
“We have defined the terms for every data piece that we’re ever going to have, so that anytime anything ever comes in — from the postal service, from another agency, from a credit card company, from another third-party entity — anything that comes in, all of our ‘datatorial’ people then cleanse it like that any they put it in that structured format. Once it’s all structured or standardized, then we can play with it.”