UPS turns data analysis into big savings

When you have an organization as large as UPS, with its armada of tens of thousands of brown trucks prowling the streets of the world to pick up and drop off packages, even the smallest inefficiencies can add up to substantial costs.

The reverse is true, as well, of course, and that's why the company is in the middle of a long-term technology upgrade for its giant fleet of delivery vehicles that is designed to wring every possible efficiency out of the trucks. In the process, it will save millions of dollars a year, make a significant reduction in the company's carbon footprint and keep its thousands of drivers a little bit safer.

And it all comes from the collection and analysis of vast amounts of telemetry data gathered from the smallest sensors found just about everywhere on those ubiquitous brown trucks.

In the United States alone, UPS puts 88,000 drivers out onto the streets every day and about 102,000 a day globally, said Jack Levis, UPS' director of process management. All told, that adds up to about 3.3 billion road miles a year worldwide. And over that massive system, the company's telemetry program, which it began instituting in 2008 and has since rolled out across about 10,000 vehicles in the U.S., is already saving the company millions of dollars a year. That savings should only grow, as the number of vehicles in the telemetry program will double in the next year.

The obvious thing to have when looking at how to ensure the maximum number of trucks are operational at any one time is a structured program of preventative maintenance. And for decades, that's just what UPS has had in place, said Levis. The rationale has long been that it's a lot cheaper to proactively repair, say, a starter every two years, than to have the part fail and keep the truck in the shop.

And that's a fine approach, Levis said, except even that proactive maintenance philosophy leaves plenty of room for additional efficiency since, for example, a truck may well go more than two years without its starter failing, especially if it's used on a light route. Conversely, a heavily used truck might see its starter die much sooner. So why treat two trucks with very different usage patterns the same, the argument goes, when doing so leaves a lot of money on the table at the scale of a company like UPS.

"That was the genesis" of the telemetry program, Levis said: "How to gather data and use advanced analytics. By using telematics, can the data, with advanced analytics, point to relationships in the data that's not readily apparent without" it?

On the surface, it sounds like a simple problem: figure out how the fleet is performing and look deep into the data to see how to best maintain the vehicles. But in practice, UPS presents a computing problem that it says few, if any, other organizations on the planet have found a way to solve.

That's why the company keeps a small stable of folks like Levis around, mathematicians and Ph.D.s who are trained to apply complex analytical models to the mountain of data coming in every day and to look deep into the numbers to find the kinds of relationships that will help UPS perform at peak efficiency.

To Levis, building these computing models on top of the mass of data being sent every day from the fleet is tantamount to giving the company's mechanics a clairvoyant advantage over the laws of physics that govern how long the trucks can run before breaking down.

'You have to mine it'
"There's millions and billions of pieces of information flooding out of the vehicle [at all times] into this dumb telematics device," Levis said, adding that as the data comes from sensors in the vehicles' engines and other areas, they are moved into big databases and merged there with other data UPS keeps on its drivers, its vehicles, and its customers' packages.

"Now you have to mine it," Levis said.

One place this data mining begins is when the company's mechanics plug in information from a truck that has failed in some way and compare its performance to that of every other vehicle it owns of the same make and year. The idea is to look for precursors in the data from the sensors of the broken down truck and compare them to data from the same sensors in the other vehicles. By identifying what was different in the data in the failed vehicle from the others, it may be possible to identify a precursor to a maintenance problem that can be searched for in the data that's coming in from the rest of the identical members of the fleet.

You would think that this approach would only apply to vehicles that have been on the road awhile, but Levis said UPS has used it to test out new trucks as well. Not long ago, it ran a test of 500 new members of the fleet and the telematics system revealed that a few of them had come straight from the manufacturer with dysfunctional parts.

Ultimately, of course, what makes this telemetry process worthwhile is the analytical systems that can figure out what the various data mean. It's one thing to plug in a bunch of sensors on a UPS truck and return a bunch of data, but, as Levis said, "You don't know what it means, so you have to do the data mining to figure [it out]. We plug it in, and then go off looking for another nugget, another piece of information [and] another relationship we didn't know existed."

The results aren't the sexiest in the world, but they can add up to major savings, Levis suggested. For example, he said, a company mechanic was recently working on a fuel injection system failure on one truck, and ended up doing a $2,000 engine overhaul. But afterward, when evaluating the data coming from sensors inside the engine, the company was able to determine that, in fact, the problem was little more than a bad O-ring. And by applying that knowledge from now on, UPS can save a lot of money.

Similarly, UPS looked at problems with dead truck batteries. By evaluating the voltage data coming in from the batteries, it was able to determine that in many cases, drivers were simply leaving the trucks' lights on. And by realizing that the problem was something so simple, the company is able to be proactive and design in warnings if the batteries drop too low.

Green impact
Levis said another big goal of UPS' telemetry program is to cut the company's impact on the environment, both for altruistic reasons, and for the economic benefits that come with doing so.

The biggest example of how this is working is found in the aggregate daily idling time of the thousands of trucks in the UPS fleet, Levis explained. By mining the data, the company was able to determine that drivers tended to leave their trucks idling for about 15 minutes longer per day than is necessary. According to the company, that equates to 25 gallons of fuel per driver per year, and about 1.4 million gallons per year in the United States alone.

Cutting the idling time, then, saves fuel and reduces emissions and wear and tear on the trucks.

"So you're getting a triple benefit just by the knowledge of when a vehicle's left on when it didn't need to be left on," Levis said.

Having that knowledge, of course, requires implementing a behavioral change among the thousands of UPS drivers, but Levis said that part has been easy. Perhaps drivers wouldn't initially believe that they were leaving their trucks idling for so long each day, but presented with the data, they have quickly adapted, he said.

There are also important safety implications emerging from the program. According to Levis, the telemetry data has been able to show UPS how often the drivers use their seat belts and how frequently and where its drivers put their trucks in reverse. Putting trucks in reverse is frowned upon, especially in residential areas, because it is the leading cause of accidents.

As with the case of idling, the drivers may at first have been surprised to be shown data demonstrating how often they back up or don't wear their seat belts. But changing that behavior has been simple, Levis said, because the drivers have a stake in being safe. Asked if the drivers have balked at their privacy being invaded by the telemetry systems, UPS spokesperson Dan McMackin said, in fact, the company's comprehensive health and safety committees--made up of drivers--"love having a piece of paper" showing ways they can improve safety.

As a result, the company's data show that drivers have improved their seat belt usage from an already impressive rate of 98 percent to 99.8 percent, and that they have reduced incidents of backing up their trucks by about 25 percent.

Ultimately, UPS feels that the data analysis it does with the information coming in from its thousands of trucks gives it a significant competitive advantage. McMackin said that as far as he knows, companies like FedEx have not developed such complex systems.

FedEx did not respond to a request for comment.

"You wouldn't know what it meant," Levis said of the mountain of raw data that comes from the trucks. "Our analytics is the magic."