Follow Datanami:
October 26, 2015

Why Big Data Is a ‘How’ at UPS, Not a ‘What’

“Take care of little things and big things will take care of themselves.” That was the advice shared by UPS’ Senior Director of Process Management Jack Levis, who presented a keynote Monday at SAS Analytics 2015 conference about the company’s decades-long journey from descriptive to prescriptive analytics.

Over a 40-year career at UPS, Levis has overseen the implementation of various analytics tools and techniques in the name of improving the massive logistics operation at UPS, the 108-year-old global shipping giant that ships more than 15 million packages per day.

The analytics journey of the Sandy Springs, Georgia company started back in the 1990s, when it implemented a suite of solutions that gave it basic descriptive analytics capabilities.

“We took data from many sources,” Levis said, including handheld computers that gather information from packages, telematics devices installed on trucks, and GPS and GIS systems used by planners. “We put that together and we found that we could find great insight into what happened yesterday.”

As the scale of the company ramped up at the turn of the century, UPS realized that descriptive analytics only got it so far. In order to take its efficiency to the next level, it would need to enhance its package tracking capabilities, and that meant getting rid of its decentralized data architecture. “Some of the data was in corporate repositories, some of it in local repositories, some of it spreadsheets, some in people’s heads,” he said. “So in 2000 we built a centralized data model.”

The new system, combined with a proliferation of even smarter handheld devices carried by drivers, had a profound impact on UPS’s operations. According to Levis, the new suite of data tools allowed UPS to reduce the number of miles drive per year by 85 million, which eliminated 8 million gallons of fuel and eliminated 85,000 metric tons of carbon dioxide.

“This got us from descriptive to predictive,” Levis said. “When you ship a package we start predicting when it’s going to get to its destination.”

Today the company is moving into the world of prescriptive analytics, which Levis sees as the ultimate destination on the analytics journey. The implementation of the company’s ORION system will automate much of the process of ensuring that the right package gets on the right vehicle on the right day.


UPS routing — Courtesy: NOVA

There are three to four components of ORION, according to Levis, including a big focus on wringing even greater efficiencies from the routing of delivery trucks. “You need to understand the data limitations,” he said. “For us, maps were a big source of limitation. To make our algorithms work, we had to make maps that no one else had.”

The average UPS driver makes about 120 to 140 stops per day. If you do the math (which Levis’ team did), that comes out to a huge number of possible combinations of stops–something on the order of 6.8 x 10 to the 60th power.

“We don’t realize the complexity that we’re under,” he said. “There’s UPS drivers who have delivered 120 stops for 30 years and they can’t possibly think through all these turns. But the prescriptive analytics can. They can find things that our people can’t on the road.”

When ORION is fully implemented in mid-2016, it should allow UPS to reduce the number of miles driven by 100 million miles, saving 100 million gallons of fuel and eliminating 100,000 metric tons of carbon dioxide. It should save UPS $300 million to $400 million per year, Levis said.

The ORION implementation at UPS provides a case study on how to move to prescriptive analytics. While it’s not easy and few companies are doing it, it has to be the ultimate goal, Levis said.

“Organizations can’t stop at descriptive analytics, and you can’t stop at predictive analytics,” he said. “If that’s where your vision ends, you’re leaving money on the table. You have to move all the way through this continuum…Gartner says there’s only a few organizations using analytics in the area of predictive and prescriptive. So your organization really needs to figure out how to be in that elite group of companies….”

Succeeding in prescriptive analytics will require laser-like focus on process improvement, and not being distracted by big data hype. That is easier said than done, but UPS and Levis’ group provide a model for how to accomplish that.

“Everybody’s talking about big data and everybody’s chasing big data and honestly I don’t have a single big data problem. I have business projects,” Levis said. “Insight that doesn’t lead to a better decision is trivial, so you need to think through what’s the decision you’re going to make with data.”

Like the dog who chases the car, beware of chasing big data without thinking it all the way through. “When you think of big data, you should think of what you’re going to do with it once you capture it,” Levis advised. “It’s the business decision that matters. Big data is a how. It’s not a what.”

Related Items:

One Deceptively Simple Secret for Data Lake Success

The Future of Data Science

UPS Delivers on Prescriptive Analytics