Next time you see one of those ubiquitous brown UPS trucks trundling along, think about this tidbit one company executive recently shared with me: That driver has more ways to run his delivery route than the number of nanoseconds the earth has been in existence.
"That's how many different alternatives there are," Jack Levis, director of process management at UPS, told me.
If that's not mindboggling enough, now consider this: Using prescriptive modeling, UPS has figured out a way to optimize each and every one of those delivery routes -- turn by turn, doorstep to doorstep, day by day. Did I mention UPS has about 55,000 package car drivers in the US (106,000 drivers globally for its entire vehicle fleet) and delivers more than 16 million packages daily?
"The optimization will find what it thinks is the best way to deliver the route today while still meeting all the business rules -- and it'll spit out that answer in six to eight seconds," Levis said.
This is not about finding big changes in how a driver should run a route. In fact, if you eyeballed the optimized route, it'd look awfully similar to a driver's normal route. That's because UPS long ago found the "real big things and low-hanging fruit" by applying descriptive analytics to its routes and then fine-tuning even further using predictive analytics.
When you scrutinize the optimized route, what you may see is a quarter mile shaved off here or a half mile there, Levis said:
The general path is good, but the optimization has made some very specific changes based on the exact packages for today. So normally you deliver House A, then B, then C, and now it says, 'Well, there are other different addresses today, so it's better to deliver House C, B, and then A because we found a way to do the reverse and get some savings.'
The prescriptive model "will find minute savings throughout the day by focusing on an early change that impacts something later. It's looking to reduce inefficient time or backtracking," Levis added.
UPS has long been analytically inclined, an analytical innovator to be sure. But this particular project is of such unprecedented scope that you wouldn't be wrong to consider UPS in an analytical class of its own. Let's explore.
Page 2: Prescriptive Analytics to the Extreme
Page 3: The Long Haul to Implementation
Page 4: Driving Change Takes Collaboration
Next Page: Prescriptive Analytics to the Extreme