http://www.informationweek.com/shared/printableArticleSrc.jhtml?articleID=17
602194

UPS slashed the time it takes to determine the least-expensive route from
months to days to hours and wants to make that information available in real
time

By Beth Bacheldor,  InformationWeek
Feb. 9, 2004
URL: 
http://www.informationweek.com/story/showArticle.jhtml?articleID=17602194

In the mid-1990s, United Parcel Service Inc. decided it needed a software
program to map its entire U.S. operations‹every pickup and delivery center
and every sorting facility, totaling more than 1,500 locations‹to find the
best routes to move more than 10 million packages and documents daily. But
there was a hitch. "It was thought to be impossible to model the UPS network
and optimize it," says Jack Levis, project portfolio manager with the
company's industrial engineering division.

Impossible because there are 15.5 trillion options for calculating every
possible route among just 25 points‹a job that would take the fastest
computer in existence 500,000 years to compute. So a group of UPS
researchers set about creating optimization software that would assess only
the most-feasible options and over the years whittled down the time to
figure out the least-expensive and shortest U.S. delivery routes from months
to just hours today.

Now, almost a decade later, UPS not only uses that homegrown software to
help answer questions such as where to put a new distribution center, but
it's planning by 2006 to give its 70,000 drivers unprecedented tools to
fine-tune the supply chain. If successful, the plan would change
supply-chain optimization from today's top-down, management-driven approach
to one that's essentially bottom-up, giving more information to the people
closest to the customer. UPS is investing $600 million in the initiative,
which includes state-of-the-art handhelds and new software and is expected
to save the company $600 million annually.

    

Operations research has become more important to UPS, Levis says (foreground
with Mohr on left and other team members).

Photo by David Deal
UPS says it has already reaped numerous payoffs from the millions of dollars
it spent on operations research that led it to build the proprietary ground
and air supply-chain-optimization technology, including saving hundreds of
millions of dollars on its air deliveries. In the next two years, it expects
more benefits, by giving drivers access to data from the
supply-chain-optimization models in real time via wireless handhelds. That
will let them adjust delivery schedules on the fly to deal with everyday
problems such as traffic and bad weather.

UPS's efforts are key in a fiercely competitive market. It holds the lion's
share of the ground-delivery market, but FedEx Corp., a company about
two-thirds the size of UPS with $33.5 billion a year in revenue, leads in
overnight deliveries. UPS and FedEx constantly one-up each other in
cost-cutting, service improvements, and new business ventures. (For more on
FedEx, see Time To Deliver, Jan. 12.)

Also a growing threat to UPS is DHL International Ltd., the leader in
cross-border express deliveries. It's making a big play for the U.S.
ground-delivery market, acquiring Airborne Inc.'s ground operations last
August for about $1.1 billion.

In this ultracompetitive environment, UPS has revised its company charter to
cite as its mission "synchronizing global commerce." That means helping
customers manage and coordinate the flow of goods, information, and money
throughout their supply chains. UPS for years has been using operations
research, a discipline that leverages mathematical algorithms to predict and
help improve the behavior of complex systems involving people, machines,
materials, and procedures. But the company's changing mission, which
increasingly encompasses global service across more than 200 countries, has
raised the importance of operations research. "There are 50 people doing
operations research here, and they're doing things that have never been done
before, that no one has ever accomplished," Levis says. "UPS's vision of
synchronizing commerce means that operations research becomes more and more
important as our customers ask us to do more and offer more specialized
services, and to be more integrated into their businesses." For example, UPS
has developed Web-based software for DaimlerChrysler AG so the automaker can
centrally manage all parts moving to and from more than 4,500 dealerships.

UPS is out front in developing its own high-powered systems, but more
companies are showing interest in optimization software. That's because
computers can now handle the massive number crunching that optimization
requires, and there's a lot more usable data‹generated by applications such
as enterprise-resource-planning or forecasting systems‹that can be used in
an optimization program, says David Simchi-Levi, professor of engineering
systems at MIT and co-author of Managing The Supply Chain (McGraw-Hill,
2003). That's led to more applications being written to solve specific
business problems.

UPS chairman and CEO Michael Eskew founded the company's corporate
industrial-engineering operations research division in the mid-1980s. "It's
vital that we manage our networks around the world the best way that we
can," Eskew says. "When things don't go exactly the way we expected because
volume changes or weather gets in the way, we have to think of the best ways
to recover and still keep our service levels."

When first presented with the challenge of modeling UPS's ground network in
software, operations research development manager Doug Mohr and his team
near Baltimore tried to optimize a subset of the network. They quickly
learned the network was too interconnected to tackle piecemeal. The team
applied methodologies it had been refining for years, along with heuristics
(a search procedure that attempts to find optimal solutions to problems),
and ran that against mathematical equations representing business
information (such as package volume and number of sorting hubs) as well as
the company's business rules, and ultimately achieved a model of the entire
UPS network that could be calculated on a Sun Microsystems Unix server. The
problem was that the optimization calculations still took at least 90 days.

It took years‹plus faster computer hardware, lots of trial and error, and
researchers constantly improving their algorithms‹for the optimization
software's computation time to drop, first to 30 days, then five, and
finally, by late 2001, to hours. Today, the program can run in less than
three hours.

During the same period, and in much the same way, the other half of the
operations research group, in Louisville, Ky., partnered with MIT in an
effort to develop optimization software that would help it improve
air-delivery operations. UPS rolled out that system, which cost about $2.3
million, in February 2000.

"Both our air and ground networks are as cost-competitive as any in the
industry," boasts Larry Subacz, industrial-engineering manager for UPS's air
group, who's responsible for network planning for domestic and interna-
tional airlines and facilities. "Without the capability to look at some of
the options we've been able to look at in the past few years, I don't think
I could have said that." The ground- and air-network-optimization systems
leverage the real-time information coming in through UPS's extensive
package-tracking system. That includes bar-code labels containing all the
data the company needs to deliver packages on time. Wireless handhelds
synchronized with UPS's back-end databases as well as state-of-the-art
conveyor, sorting, and scanning systems also play a role. The tracking
system incorporates new logistics software, being developed at a cost of $20
million, that will aggregate ZIP code information and map out how packages
should be loaded onto trucks for the most efficient deliveries.


    

UPS's Baltimore sorting operation is one of 1,500 U.S. facilities.

Photo by David Deal
The client-server system for optimizing the hub operations and delivery
fleets of its ground network, called Hub and Feeder Network Optimization, is
in the hands of five full-time planners. They set up problems on PCs by
entering data from point-to-point volume files, which outline package
origination, destination, and volume, and are built by various forecasting
applications; data from sorting facilities' applications, which provide
basics such as location and capacity, plus time and distance between sites;
and data from a flow file that defines the paths packages take between
different sortings. Once the problem is set up, it's transmitted to a Unix
server, which runs the computations and downloads them back to the PC as
reports.

Hub and feeder network optimizations are typically done only a few times a
year, and the answers they provide help UPS plan for five, 10, even 20 years
out. "We're trying to answer things like, when will we run out of capacity
with our current network? Where should we expand? What is the impact of
changing our service level to our ground operations? What about changing
costs?" Mohr says.

There's an even more-ambitious goal: optimizing in real time. By 2006, UPS
wants to use the software on the wireless handhelds so its thousands of
drivers, who each average more than a hundred stops a day, can optimize
their own routes in minutes within the context of UPS's overall operations.
"That's where the magic happens," Levis says. "We will have merged all the
demands‹information about each package with information about customers'
needs, wants, and demands with information from maps‹and we'll put it in the
hands of the drivers."

UPS is testing early iterations of the system on the newest generation of
its custom-developed handhelds, which incorporate local and wide area
wireless connectivity, a Bluetooth link to communicate with peripheral
devices and PCs, and links to global-positioning satellites. The handhelds'
software will be synchronized with PC-based dispatch-planning-optimization
systems being deployed to 1,600 delivery centers around the country. Those
systems will churn data constantly, including all drivers' delivery
schedules, package information, and up-to-date maps. UPS hopes its drivers
will be able to juggle time conflicts on the fly, such as getting a certain
package delivered first to meet a deadline while a raging snowstorm shuts
down a major roadway.

That's more evidence of UPS's pioneering spirit. "Almost everything UPS does
now wasn't even conceivable five years ago," says Michael Trick, professor
of operations research at Carnegie Mellon University and past president of
the Institute for Operations Research and the Management Sciences, a
professional organization. "They are definitely one of the leaders in taking
the state of the art in optimization and expanding its boundaries, doing
things faster, and solving bigger problems."

Five hundred miles from the nondescript office building in a Baltimore
suburb where employees, including Mohr and his team, ply their trade, you'll
find the other 25 members of the optimization group. They're in Louisville,
home of Worldport, UPS's 4 million-square-foot automated package-sorting
facility, which can sort 304,000 packages per hour, and air hub, which
opened in fall 2002. The system used for optimizing air operations at
Worldport, called Volcano (which stands for volume, location, and aircraft
network system), takes into account all the delivery routes for the 265
UPS-owned and 316 chartered aircraft. It delivers answers about optimized
routes, fleet assignments, and package allocations in days or even hours.
That's important because, according to pilot-scheduling work rules agreed
upon by UPS and the pilots' union, UPS must update its air-operation
schedules at least 45 days before new ones go into effect, and it does 13
such updates each year. UPS executives expect the system to save the company
more than $200 million over the next 10 years by helping it better plan and
schedule air deliveries, which total as many as 2 million packages and
documents each day in the United States alone.

Next, the team is building an air-optimization system for packages that come
through the air network sorting facilities and travel on the aircraft. It's
still a relatively early prototype, says Keith Ware, manager of the air
operations group in Louisville, whose team is also trying to develop a
system that would produce optimal aircraft rotations by considering both
next-day and second-day services simultaneously.

UPS planners also consult Volcano when deciding whether to bring on new
aircraft. Shortly after Volcano went live, the planners in charge of
preparing for peak season used it to determine whether the company should
lease additional higher-capacity aircraft or find another way to handle the
increased shipping volume between Thanksgiving and Christmas. Daily delivery
volumes can more than double during peak times‹on Dec. 23, UPS air express
deliveries can top 4 million, twice that of an average day, and the number
of flights that take off the week before Christmas can jump 45% to more than
1,300‹so routing and capacity decisions are critical. A misstep could put
deliveries in jeopardy or cost the company millions of dollars more than
necessary for last-minute flight additions. The system calculated that by
changing the routes of some flights in the existing fleet, UPS could get by
leasing just one more aircraft instead of several planes that could have
added up to $3 million in costs.

UPS planners tap optimization applications‹including those for ground and
air routes, plus others for its tractor-trailer and delivery fleets‹as the
company expands its footprint throughout the world. Last year, the team ran
1,500 optimizations to help determine where and how UPS should build out its
European operations. Compare that with 1986: When UPS was planning to put a
new facility in the Chicago area, it took four people three months to run
one such calculation. "Today you can make a much more fine-tuned decision
with few people," Levis says.

Planners use the software to test scenarios such as the best location for a
new sorting facility in Germany or how big that facility should be. "For
every one of those questions, you need to know how much each answer will
cost," Levis says. There's precedent. UPS used the software in 1999 when the
company decided to expand the Louisville facility into the Worldport hub.
"That was a billion-dollar decision," Levis says, referring to the $1
billion price tag for Worldport. UPS had considered multiple options,
including building a new hub elsewhere or modifying volumes through regional
hubs, but the software showed a Louisville expansion made the best sense,
considering costs, how long it would take to complete the job, future
growth, and the makeup of the air fleet, among other things. "The answer
always generates more questions, and you can compare quantitatively which
idea was better," Levis says.

Relying too much on instinct can be costly. Levis says a VP once approached
him with an idea for a new building as part of a plan to redo where and how
packages flowed through UPS's sorting facilities. "Intuitively, it made
sense to me," he says. "But we decided to model the problem in the system,
and when we did that, it put almost no packages in the building." The
software showed the building wasn't needed and avoided what could have been
a $100 million-plus mistake.

UPS's optimization software regularly helps improve service to customers.
Just recently, UPS cut at least a day off guaranteed delivery times for
certain shipments. For example, shipments between Los Angeles and New York
are now guaranteed to arrive in four business days instead of five. UPS
credits changes to railroad service for some of the improvement, but it also
modified its interstate trucking network and package-sorting times and
locations with the help of the optimization tools.

UPS's aim is to stay at the forefront of supply-chain optimization.
"Optimization has allowed us to make each transaction and each customer a
one-to-one experience," CEO Eskew says. "It's going to keep our service the
best in business today."


--~--~---------~--~----~------------~-------~--~----~

TELECOM-CITIES
Current searchable archives (Feb. 1, 2006 to present) at 
http://www.mail-archive.com/[email protected]/
Old searchble archives at 
http://www.mail-archive.com/[email protected]/
-~----------~----~----~----~------~----~------~--~---

Reply via email to