Two years ago, Toyota Motor Sales USA Inc. decided that its U.S. distribution network was due for a tune-up. The three-decade-old system of warehouses had been established at a time when the Japanese automaker sourced most of its parts from overseas to serve a small network of U.S. dealerships. But that scenario changed in the '90s when Toyota shifted more of its business to North American parts suppliers and its dealership network exploded.

Under the existing system, the Torrance, Calif.-based company has been providing U.S. after-sales support to 1,200 car dealerships, 200 Lexus luxury car dealers, and 100 forklift dealers via a two-tiered system. The first tier consists of two large distribution centers (DCs)—one in Ontario, Calif.; the other in Hebron, Ky. Those two sites, in turn, feed parts to nine smaller sites located around the country—in Los Angeles; San Francisco; Portland, Ore.; Kansas City, Mo.; New York; Cincinnati; Baltimore; Chicago; and Boston. The company also operates a facility strictly for Lexus parts in Jacksonville, Fla.

Toyota had not undertaken a strategic network analysis since 1978. But its operation has changed significantly since that time. For starters, its customer base has grown. It also sources differently today, bringing in 55 percent of its parts from North American suppliers rather than from Japan. Finally, in addition to supporting its Toyota models, the company has added parts distribution for its Lexus line of luxury automobiles, which were first introduced in 1989. "The decision to go through with a network analysis/simulation was strategic," says Susan Dexter, a business process change manager at Toyota who oversaw the project. "We wanted to be proactive and make sure that we could continue our high levels of customer service in light of our projected growth over the next three to five years."

But what would be the optimal network for an organization that moves more than eight million parts and accessories around the country each month? To answer that question, Toyota turned to computer modeling. Using network simulation software, the automaker decided it would first examine the distribution network used for its Lexus division and then look at the entire network. "We wanted to do a comprehensive study of our DCs to see if they were in the right place to meet the dealers' needs," says Dexter. "Our objective was to develop a parts logistics network to support business growth and maximize customer satisfaction," she adds. "If we could save a few dollars, that would be great, too."

The results of the software modeling revealed that Toyota could improve customer service to dealers while cutting costs by opening a new distribution center. Despite the startup costs, the study showed, a new DC that would strictly handle Lexus parts would quickly pay for itself by eliminating the need for premium-priced expedited transportation and also alleviate overcrowding at an existing distribution center. On top of that, the model indicated that customer service could be improved with faster delivery.

The modeling exercise began in December 1998, when Dexter was put in charge of the network-design project. She quickly assembled a team of 10 full-time employees from all over the organization to work with her.

One of Dexter's first decisions was to use a software tool called "SAILS" from Insight, a Manassas, Va.-based software firm. SAILS software, which models logistics networks and provides analysis, is designed to help companies achieve their customer-service objectives at the lowest cost.

Before the SAILS software could be put to work, however, the team had to feed large amounts of data into the application to develop a model that reflected Toyota's existing distribution network. "The first thing you do with SAILS is develop a baseline," Dexter explains. "You run all your demand [data] through a model of your network the way it's configured today. For instance, a dealer in Cincinnati would order from the Cincinnati PDC (Parts Distribution Center) and you'd capture all the costs. Then you'd go back to your historic data and say, 'Does this make sense'? And you'd make sure the costs lined up and tweak the model."

Data categorization and collection proved to be a considerable task for the team, consuming five of the six months devoted to the project's initial phase, which looked strictly at Lexus parts distribution. In that first phase alone, Dexter says, the team had to gather historical demand on 250,000 active parts and feed that information into the model. "We spent a lot of time doing the baseline model," she reports. "We wanted to make sure we were pulling the right data."

The frequent need to redefine the data contributed to the slow development of that baseline model. Dexter says the team found that it had to clarify the data—deciding, for instance, what part-sale categories should be included and excluded from the model. "Originally, we had a six-month schedule for this project," she notes. "I told management that wasn't realistic." In addition, Dexter reports, her team had to work closely with the software developers to refine the application itself as they went along.

When the SAILS model was finally run for the first time, it showed that Toyota needed another distribution center in the Texas area to serve its Lexus car dealerships. "We were serving our Lexus dealers from a Kansas City parts distribution center, which was operating over its capacity," Dexter says.

The model indicated that opening an exclusive Lexus warehouse would benefit Toyota in several ways. First, the relocation of Lexus parts from the Kansas City facility would free up space in that warehouse, alleviating the need for expansion there.

More importantly, a dedicated Lexus parts distribution center would improve service and reduce costs. Dexter says the model indicated that Toyota could realize $2 million in savings by cutting back on expedited ground transportation and on the emergency air shipments of its Lexus parts.

On top of that, a dedicated facility would improve customer service to Lexus dealerships by speeding up parts delivery. At the time the study was run, Dexter reports, the Kansas City distribution center was shipping 40 percent of its parts within one day from the time the order was received and the remaining 60 percent in two days. The model showed that a dedicated facility would be able to ship 88 percent of the parts orders in one day and the remaining 12 percent within two days.

Toyota later ran the model for its full distribution system. Although the computer modeling backed up the earlier recommendation that the company open another warehouse for Lexus parts, Toyota learned some other things from it as well. Dexter says the model forced her to reflect on the nature of a good stocking policy to serve customer expectations. She further notes that the project has started Toyota thinking that customers with newer cars might have different parts-replacement service expectations from those of customers owning older vehicles.

Since the project ended in December 1999, Toyota has begun the process of siting a new parts center. The company has settled on Texas as the location of the proposed facility, which it hopes to have up and running later this year. The automaker has not yet determined whether it will operate the proposed 60,000-square-foot facility itself or outsource the operation to a third-party logistics concern.

In addition, the modeling exercise has opened up Toyota executives' eyes to the possible benefits of reconfiguring the company's U.S. distribution network. "There will be an improvement in customer satisfaction from being able to deliver parts more quickly from the distribution center," says Dexter. "The increase in our customer service and responsiveness will be significantly enhanced—that was our project's selling point."