Kristine Schmidt plays a lot of roles depending on what the day has in
store for her: project manager, logistics and operations analyst,
information-systems expert, or interpreter between top Waste Management
Inc. executives and the folks driving the company's trash collection
trucks.
As fleet optimization manager for the New Jersey market area,
her job is to free colleagues such as Steven Masterson, senior district
manager for Waste Management of Delaware, from using highlighters,
string, pushpins and maps cut into pieces and then pasted together to
create routes for his drivers. Those were Masterson's tools until early
last year.
"Our system was antiquated," says Masterson. "By the time we
were finished, six months to a year later the routes had to change [due
to new customers and acquisitions]."
Enter Schmidt. Her goal was to centralize the most basic, and
arguably the most important, process the trash-hauling giant faces
every day: the sequence of stops each truck must make to pick up the
daily refuse of American life.
It's no small issue. At the end of 2003, the company operated
18,850 routes in every state in the U.S. If Waste Management can
eliminate one route, which includes a truck, a driver, fuel and
maintenance, it can save $10,000 a month. That's $120,000 a year per
route. In 2003, Waste Management was able to save $18 million and is on
track to save $44 million in 2004. That's a boost of 7% for a company
that earned $630 million in 2003. Over a five-year period, Waste
Management plans to save $498 million on operations, resulting in a
cash flow increase of $648 million.
Not bad for a $10 million initial investment in a homegrown application it now calls WasteRoute.
"This is a moving process," Schmidt says, without irony. Each
of the company's 66 market areas—with a geographic region such as New
Jersey and Delaware comprising an area—includes an average of 285
routes. To reroute the trucks in a given district took six to 12 months
five years ago; now, it takes four weeks. If the company can map the
route system efficiently, the savings can last a year or more. "We're
trying to be proactive so we don't have to do a complete reroute every
six months," she says.
Waste Management relies on operations research, a quantitative
technique designed to create models to predict the behavior of a system
and the humans that operate it. Operations research uses mathematical
theory, statistics and computing to solve problems. In Waste
Management's case, mathematics is applied to the numbers behind pickups
such as the time it takes a driver to hoist a container via a robotic
arm versus doing the job manually, or the time needed to deal with
construction detours.
Once dozens of variables are combined into a route knowledge
bank, WasteRoute uses algorithms to pick the most efficient paths.
These algorithms sort through the myriad issues Waste Management
executives face daily: Is construction delaying a driver? Are there
promised windows of time to pick up waste? Is a driver close to running
more than the 60-hour-per-week maximum imposed by the Department of
Transportation?
WasteRoute's rollout took 18 months to plan. First, Waste
Management tried multiple off-the-shelf applications before determining
it needed to build the application itself to account for all the
variables it faces, such as when a local ordinance prevents a truck
from passing over a storm drain. A Woodlands, Texas, logistics software
company, Institute of Information Technology (IIT), won a 19-vendor
bake-off to figure out the mathematical formulas for such factors as
estimated pickup time, which side of the street to be on, and truck
speeds. IIT won a final three-supplier round after Waste Management
figured its algorithms were the best at cutting routes, balancing
workload and meeting constraints such as a 400-customer cap per
residential route.
IIT's marching orders were to develop a routing system that
could be used by any Waste Management executive or manager via a Web
browser. Drivers still get information via printouts called route
sheets.
The result is delivered over Waste Management's intranet. The
application pulls data from the company's sales, customer service and
operations databases, plots addresses to a map and then recommends
route changes to increase efficiency. For instance, if a commercial
customer gets three pickups a week for one 4-cubic-yard container, it
may be possible to "upsize" him to an 8-cubic-yard container and cut
visits to once a week.
Cleaning Up Internal Waste
The WasteRoute project is part of Waste Management's effort to
instill management rigor to a company that in the 1990s was muddled by
hundreds of acquisitions—one a day on average in 1999—shaky accounting
practices, weak internal controls and investigations by state attorneys
general. In 1998, No. 3 player USA Waste bought No. 1 Waste Management
when the larger company was facing financial ruin, Securities and
Exchange Commission probes and class-action lawsuits seeking
undisclosed damages. USA Waste then adopted Waste Management's name and
promptly stumbled, cutting earnings estimates three times in July and
August 1999 for the quarter ended June 30, 1999. The class-action
lawsuits soon followed; Waste Management settled them in November 2001
for $457 million.
In 1999, Waste Management began centralizing the company and
revamping its management team. The goal: Give two decentralized
businesses a common objective. In 2002, Waste Management collapsed more
than 1,200 operating sites into what are now 66 market areas such as
the New Jersey/Delaware region and Northern California. It also
consolidated its information systems. Before the restructuring, daily
operations were conducted with customized IBM AS/400 software
distributed to more than 400 remote servers. Today, Waste Management
operates under a consolidated, internally developed system for
accounting, billing and customer service dubbed MAS. The system is
centrally distributed to market areas and districts.
Those consolidated systems feed customer data into WasteRoute,
which then maps the information to a route. The application, which is
similar to Mapquest, can account for parameters that influence routes
monthly. For instance, when schools are in session, Waste Management
doesn't pick up schools' trash during recess because of safety
concerns. In the summer, school pickups drop from five days a week to
one or two and timing is more flexible. At the same time, pickups
increase at resorts along the shore.
When changes are needed, a manager can add parameters such as
road construction and generate a new route. Customers, meanwhile,
expect drivers to hit designated times for certain pickups. "In
Delaware, people set their watches by the mailman and the garbageman,"
Masterson says.
Alex Popov, vice president of logistics at Waste Management,
says the company is 60% done installing WasteRoute in its high-margin
commercial accounts, 30% finished with its residential services, and is
just beginning to roll out the software to its industrial customers.
Executives first tackled routes that were the easiest and most
profitable. For instance, commercial accounts—strip malls, restaurants
and small offices—are predictable because they have similar containers
and regular pickup hours. Commercial accounts also generate more income
because Waste Management hauls away more tons of waste with two or
three trips to a landfill on average compared to residential.
The wild cards include blocked containers and new customer
stops. Mileage is another issue. One driver may travel 150 miles to
pick up 60 customers where another may drive 400 miles in a rural area.
For residential waste, the biggest change was more stops,
potentially more drivers leaving cabs, and restrictions such as trash
pickups on only the right-hand side of the street. The goal is to
cluster customers in dense residential areas and reduce operating costs
by traveling fewer miles as efficiently as possible.
The third target for WasteRoute will be the company's
industrial business, which covers construction sites and manufacturers.
WasteRoute needs to be linked with data throughout the day on driver
location and incoming requests for service from the billing system and
dispatchers, since routes aren't planned in advance. Currently, Waste
Management can upload and download customer information daily to
WasteRoute, but it isn't a real-time link, according to Schmidt.
Surya Sahoo, IIT's CEO, says there will be different
constraints for the industrial business that will require a new
algorithm. A typical industrial pickup vehicle can only service one
container at a time, and 70% of the pickups are not scheduled in
advance. An industrial container is picked up, emptied at a landfill
and then brought back to the customer. Waste Management can also bring
an empty container to the customer and swap it with the full one. When
the container is emptied, the truck can go to another customer. The
problem: Customers have different container sizes and service
requirements. The new algorithm will also have to account for vehicle,
container and material types, security clearance and weather patterns,
Sahoo says.
Thow Out Those Pushpins
The industrial rollout offers more of a challenge, but Waste
Management executives are confident after recent successes.
Nevertheless, there's a reason Masterson was using pushpins to route
his drivers only two years ago. That's because previous attempts to use
routing software didn't work.
USA Waste had an application called STARs and tried to roll it
out to the newly merged company, but the software failed. It wasn't
linked with the company's other systems, offered limited visualization
and couldn't account for factors such as road construction and driver
availability.
After STARs flopped, district managers were on their own. In
early 2002, Masterson installed route optimization software for the
Delaware district. But the application, bought from a software company
led by former employees of a Waste Management rival, struggled with
optimizing dense areas, sent drivers down one-way streets and illegal
roads, and couldn't design residential routes with pickups on the right
side of the street.
In early 2003, Waste Management was determined to succeed with
WasteRoute, but needed solid data. Sahoo first took Waste Management's
customer database and mapped it to streets and locations with
information purchased from Navtech, a mapping company. But Waste
Management lacked billing information for all of its customers. In its
residential business, Waste Management is often billed by homeowner
associations and municipalities to pick up waste from a neighborhood.
One billing address could represent hundreds of homes. The company
filled in the gaps with municipal street information.
To verify data, Waste Management drivers fanned out with
Nextel phones that can capture the coordinates of any spot on Earth and
relay it by satellite to the company's databases. Drivers simply
pressed a button whenever they passed a new address. The process
confirmed the location of commercial and residential sites, and allowed
IIT to cluster addresses. "We had to validate the data and see it on a
map or it wouldn't have worked," says Sahoo.
Also entered into WasteRoute were the type and size of waste
containers—larger ones for commercial, smaller for residential—and
estimated time to serve based on driver input.
Once the data was entered and functional, Schmidt and the
managers set about selling the system to drivers. "It's human nature to
worry about change," she says.
To alleviate drivers' first concerns, Waste Management made it
clear that WasteRoute would not eliminate their jobs. The system was
cutting routes, but reallocating drivers and trucks. Any job losses
would occur through attrition, according to Popov. Waste Management
started the project with 19,600 routes and plans to end the year with
15,000. For its purposes, Waste Management considers routes and drivers
to be synonymous.
Masterson also had to sell the system on the front lines. One
18-year veteran told Masterson he would quit if he had to go through a
route change. Masterson walked the driver through the changes verbally
and on a map. He even brought the driver's wife into the loop.
After the drivers were convinced that the change was for the
better, Schmidt began to show them computer demonstrations. As she
recalls: "It was a great feeling when I heard, 'That's pretty cool, can
I see my route?'"
