Follow Datanami:
February 23, 2013

UPS Delivers on Prescriptive Analytics

Nicole Hemsoth

When it comes to complex IT operations, few Fortune 500 businesses rival the complex logistics challenges a company like United Parcel Service (UPS) faces.

The orchestration of ordering, scheduling routes based on ever-changing conditions, package deliveries that require follow-up action, a massive fleet of air and ground vehicles reporting data, and a symphony of other requirements–these are just a few small pieces of the big data puzzle for UPS.

The package delivery titan claims to house the world’s largest DB2 relational database, which is powered between the company’s two U.S.-based primary datacenters that handle global operations. In addition to this arsenal, they are also experimenting with a Hadoop cluster and looking at ways to streamline the increasing number of data feeds from sensors, devices, vehicles, and tracking materials. This flood of new data formats into the massive existing store complicates the core logistics mission.

Keeping harmony requires world-class operational efficiency. For a company at UPS scale, this means having a fine-tuned approach to making use of the results of business analytics. While the complexities of the day-to-day business are deep, one angle of UPS operations is worth a second look—one that Jack Levis, a 36-year UPS veteran at the helm of process management, described for us in a recent chat.

Levis noted during his long tenure, he’s seen a lot of analytics trends come and go—not to mention a lot of buzz without practical cause. He lumped “big data” into that trend list of things that fizzle, even though he says that data element is a standard struggle for all large companies, just as it always has been.

“We hear so much about big data now, but really, UPS has been in the business of big data for a long time now,” says Levis. His goal is to use bigger picture analytics to analyze operational efficiency, spelling out how descriptive, predictive and prescriptive analytics power their daily operations. While much has been written about the models, tools and approaches behind this big data triad, it’s clearer to put the value of business analytics in practical context.

Levis argues that where data-driven businesses derive their strength is the point just past both descriptive and prescriptive analytics. He says that descriptive and predictive analytics help the carrier understand what they did yesterday to better predict what will happen tomorrow, the highest value comes from prescriptive analytics. This approach takes prediction one step further by letting companies set their roadmaps to reflect the past, present and future as a whole to optimize all operations and even understand how optimizations will affect all parts of the organization. . Levis says prescriptive models help companies like UPS more cleverly wrap new strategies around existing processes.

Jai Menon, IBM Fellow and CTO for Technical Strategy within IBM’s Systems and Technology group notes that there is so much data that requires filtering that what businesses want it to move beyond predictive models and into prescriptive ones. “It’s not just telling you what’s going on, it’s not just telling you what might happen, it’s also telling you what to do about it—this is what’s going to happen, so here’s what you need to do.”

While descriptive analytics show the rear-view mirror and predictive analytics help UPS forecast or model possibilities, the prescriptive approach ties it all together to help improve overall operations. “Descriptive analytics are easy,” says Levis. He says that predictive models are increasingly complex, but the real innovation in terms of algorithms and modeling is reserved for big picture operational view prescriptive analytics provides.

While Levis told us that real-time analytics aren’t a large part of the day-to-day delivery options, he said that there is a great deal of data that is processed in batch quickly based on everything from driver activity to the moment a package starts its tracking journey. All of these data feeds (GPS, driver route and workloads, traffic patterns and delays, drivers speeding or not wearing their seat belts, customer delivery requirements, etc.) create a constantly-changing operational landscape that requires swift analytics action to optimize daily activities.

At UPS, predictive analytics are used by the frontline managers for things like planning constantly-changing driver routes and delivery schedules. Each day, the models help the management team determine a plan of what territory a driver will cover. Such decisions are powered by the “rear-view mirror” of descriptive analytics and the sophisticated models that try to assume what will happen given a certain set of route conditions.

The considerations are different for prescriptive analytics, claims Levis. “The data needs to be more accurate, more pristine for the really valuable part of what we do.” While fast processing of the wealth of data they receive is imperative, UPS is not making use of ultra-high performance systems to scurry through the wide variety of data. Further, a great deal of the analytics his side of the business works with comes from standard relational databases versus more sophisticated models for tackling big data.

“Everything has been built for exactly the purposes we’re using it for. While we’re using relational databases, these have been fine tuned for our needs” he says. While he notes that the company is experimenting with Hadoop, the focus at UPS remains on refining and tailoring the data management and analytics with the same precision upon which they base their delivery services.

These analytical approaches are at the heart of UPS’ operations—and Levis’ role at the center of them. In addition to the overall strategic and operations research required, he’s spent his decades at the delivery giant being responsible for setting the course for UPS planning, execution and control systems. It’s no small task to oversee everything from corporate pickup, delivery and transportation systems, especially with the incredible number of flows of data that stream into the company.

All of this aside, it’s worth noting that such a long career at the center of so much data hasn’t dampened his interest in the space. In addition to his role at UPS, Levis is also the VP of Practice at operations research organization, INFORMS, where he is introducing a new program to churn out Certified Analytics Professionals. For those who are interested in testing the test, the first exam will be given at the upcoming INFORMS Analytics meeting this spring in San Antonio.

Datanami