In the spring of 2006, a recreational products manufacturer approached the University of Tennessee's supply chain management program with an intriguing proposition on behalf of one of its operating divisions. The company explained that two decades of acquisitions, organic growth, and surging import/export volumes had left the division's North American distribution network in need of an overhaul. Would the school be interested in conducting a network optimization study for the division?
Our interest was piqued by the request. When faced with sweeping business change, companies typically respond by revising their manufacturing, sales, or financial strategies. They rarely think to revisit their distribution networks, even when those changing business conditions could have enormous repercussions on their distribution operations.
But here was a company that already understood the implications of business change for its network. It recognized that its distribution network would need to evolve as its business changed. It also understood the value of using optimization or modeling tools to identify the steps needed to bring the network in line with business conditions. We agreed that this was a proposition worth pursuing and invited the company to tell us more.
A call for help
The story that emerged contained elements that will be familiar to supply chain managers everywhere (as well as a few that are unique). The company, which we'll call by the fictitious name Recreational Gear Inc., was founded in the middle of the 19th century and expanded over the years to include a diverse line of consumer recreational products that today are sold throughout North America, the United Kingdom, South America, Europe, Asia, and Australia. Interestingly, this division had been divested by the company in 1970 but was re-created in the mid-1980s when the company acquired two firms in this line of business. The division continued to grow through the years with the introduction of new brands and the acquisition of several more companies. By 2006, the business division's portfolio included 29 different product categories under the management of six operating units.
This growth did not occur without some pain. In Recreational Gear's case, the pain was largely inflicted by soaring transportation costs. The company was hardly alone in that respect. By the time of its initial contact with the University, there wasn't a company in the country that had escaped the run-up in transportation costs or the pinch of the transportation capacity shortage. From 2004 to 2006, transportation costs had increased by 23 percent to an average of US $1.69 per mile. By 2006, outbound transportation costs were consuming 62 percent of every dollar spent on logistics. For beleaguered shippers, the bad news just kept coming: rate increases, large fuel surcharges, and even larger increases in accessorial fees. Worse yet, all this was happening at a time when changes to the federal Hours-of-Service (HOS) regulations had caused carrier productivity to decline.
But what really concerned Recreational Gear was that its costs were increasing by more than the national average. In fact, between 2004 and 2006, the division's outbound transportation costs per unit had risen by more than 32 percent. Inbound and other transportation costs had also increased dramatically. The company knew that its current network was contributing to its higher-than-average transportation costs.
At the same time, international imports and exports had become a growing part of the business, which brought the added complexities of moving products from manufacturing facilities to the ports and getting products from non-domestic manufacturing locations to domestic customers. The division was keenly aware that any decisions it made regarding which ports to use for exports and how imports should flow through the logistics network would have major transportation cost implications for the company.
As for domestic operations, the existing outbound network was a complex structure that essentially had a supply chain for every brand. Outbound transportation was primarily provided by dedicated third-party entities—although there was some private transportation capacity from previous acquisitions.
Furthermore, the division was using in total more than 500 specialized trailers, with an average shipment distance of about 1,100 miles. Still, there were large variations in the density of vehicle loads. For example, one plant may have been utilizing a piece of equipment 130 percent (by overlapping items in the load), while another plant was only loading the same equipment to 70 percent of the cubic capacity for the same product. Due to the nature of the specialized equipment and the lack of transportation visibility across the various units, there were also variable deadhead distances for each operating unit.
The Recreational Gear division was convinced that it could boost efficiency by conducting a network optimization study. But like many companies, it had neither the expertise in house to perform this type of study nor the necessary analytical tool. Partnering with the University would give it a cost-effective means of carrying out the study.
The benefits to Recreational Gear were obvious, but what about the University of Tennessee? We concluded that the school would benefit from the partnership as well. The University currently uses a network- optimization software package in our undergraduate and graduate logistics and supply chain program. We decided that working on a project with this company would enable us to bring current and relevant examples into the classroom, in addition to expanding our knowledge of the analytical tool.
Conducting the network analysis project
As soon as the administrative and legal departments of both organizations had given the proposal the green light, the project team began mapping out the study's various phases. As with most such projects, the first phase would involve gathering data for the analysis. The success of any network-planning project depends upon the quality of the data used in the models. In this case, the project would require a vast array of data, including a listing of all products, customer locations, stocking points and sources, demand for each product by customer location, transportation rates, warehousing costs, shipment sizes by product, customer service goals, and order patterns by frequency and size.
Because the company was using a transportation management system (TMS), we assumed at the outset that having access to a centralized data repository would make the data collection and validation stage relatively short. This ended up not being the case. The TMS was not fully operational at the time that the project started. In fact, it was operating in parallel with a legacy system. The TMS generated a tremendous amount of data that had to be cleaned and prepared for input to the optimization model. Furthermore, not all of the business units were using the TMS. As a result, the data had to be collected and compiled from multiple systems. Each business unit's distribution process had to be documented and understood in order for it to be accurately captured by the model.
Data collection is a critical step and requires the work of dedicated and knowledgeable company personnel. At Recreational Gear, however, staff members assigned to the network design project found themselves trying to juggle their regular responsibilities with their added data-gathering tasks. This led to more than a few long days and some weekends in order to complete the data collection phase.
As much work as it may be, the data collection effort is nonetheless necessary. The team needs this information in order to conduct a current-situation audit or create an "as is" baseline representation of the current network. Before any potential changes to the existing network can be analyzed, one must ensure (validate) that the model replicates the realworld situation. This is the most time-consuming—and often most frustrating—part of the project. But until the project team agrees that the model provides a reasonable representation of reality, it is not advisable to move to the next stage.
For Recreational Gear, the current-situation audit resulted in some interesting insights. As Figure 1 shows, products were being manufactured throughout the United States and Mexico and distributed across the same geography. In many cases, a driver for the dedicated carrier could literally pass one of his/her colleagues going in the opposite direction. One might be hauling outbound freight from one plant, while the other could be going the other way with freight from an entirely different plant. The freight on each might be similar products but different brands! Then each might have to deadhead to other plants to pick up additional loads. This map confirmed what the project team suspected: A great deal of efficiency could be gained from optimizing the flow of outbound products.
Making the analysis
Once the data-collection phase was complete, it was time to begin the interesting part of the project: the optimization of the baseline to identify immediate savings and benefits. The goal of this phase is to find the "low-hanging fruit," or improvements that could be gained relatively quickly and easily. An example might be incorporating the private fleets left over from acquisitions into the company's centralized transportation network. At the same time, this optimized baseline also indicates how much the current cost exceeds the lowest possible cost.
It has been our experience that optimizing the current network results in savings that get upper management's attention and generate excitement. Even though many companies would be more than happy to stop at this point and begin implementing network changes in hopes of realizing some of the projected savings, the network analysis is not yet complete. At this stage, many companies will benefit from taking some time to conduct a trend analysis. That is, the project team should determine what would happen in the future if nothing changed. "Doing nothing" is always an alternative (and a popular one at that). This analysis can be used to determine what benefits will result in the long term from making changes to the network. For example, the trend analysis can reveal the point at which manufacturing or distribution capacity will become constrained and costs and/or service will suffer unless changes are made.
Although most companies find that a trend analysis is a necessary and useful part of the project, Recreational Gear proved to be an exception. For this company, "doing nothing" was not an option. Management had already stated that it would continue to acquire companies in the same line of business. Furthermore, the company was less interested in potential transportation-capacity constraints in the future than in the immediate problem of rising transportation costs. In the end, Recreational Gear decided to dispense with the trend analysis.Whether or not it conducts a trend analysis, the project team's next step should be to create a goal statement for the network analysis. Without such a goal statement, the network design project will suffer from a lack of focus. The company must state what it wishes to achieve in the future. Examples are:
For Recreational Gear, the goal was to "optimize" transportation costs for the current network of manufacturing facilities.
Once the company has identified its objectives, it's up to the project team to state possible alternatives for achieving the goal(s). For example, if the goal is to move into new territories, the team might come up with alternatives such as building new capacity or acquiring an existing company. Listing more than one alternative will result in a greater likelihood of finding the best way.
Sometimes the alternatives for achieving the goal(s) and the immediate benefits mentioned earlier are mutually exclusive; sometimes they overlap. For example, an alternative might involve expanding a current distribution facility or opening a new facility at another location. These alternatives would require considerable time and resources and would not be considered low-hanging fruit. An alternative that would be classified as low-hanging fruit would be the merging of the private fleet with the dedicated transportation operation while implementing a network of regional hubs.
As a company lists its alternatives, it often discovers that it needs additional data. For example, one of Recreational Gear's alternatives was to create a regional distribution network. Analyzing this alternative required getting cost data for unloading and reloading product.
Weighing the alternatives
Next the optimization model is used to determine the potential savings that each of the alternatives could produce. The network-optimization project for Recreational Gear involved the analysis of five scenarios for achieving the company's goal. These scenarios are shown in Figure 2.
For scenario 1, the company would use the network-optimization tool to determine how to allocate carriers on lanes so as to move its materials and merchandise at the lowest possible cost, either by choosing the most economical carrier or by increasing the capacity utilization.
Like many companies, Recreational Gear was subject to end-of-month and end-of-quarter spikes in demand. The objective in scenario 2 was to analyze the use of seven regional hubs to buffer a period's transportation demand. Loaded trailers would be temporarily held at the regional hub until tractor units were available. Transportation savings would occur as the number of vehicles needed to cover a period's shipment was reduced. This scenario operated under two assumptions: All loaded product would clear the regional hub in three to seven days, and any loaded equipment at the regional hub would not require any reloading before moving to the final destination.
Scenario 3 involved the development of regional carrier networks domiciled at selected plants to supplement the existing long-haul network. The regional carriers would be limited to delivering out to 400 miles with empty backhauls of the same distance. It was believed that there was a potential for savings here by reducing the higher costs paid to the longhaul network for delivering short distances.
Scenario 4 involved the possibility of developing delivery hubs that would operate much like distribution centers, where workers would unload inbound shipments, remix the products, and then send them back out for delivery in a much more economical manner.
The manufacturer had several private fleets that remained from earlier acquisitions. In scenario 5, the objective was to determine the effect of integrating those private fleets into the overall transportation network. These private fleets typically carry product to customers, then return empty to the originating plant, resulting in deadhead miles equal to the loaded miles. In contrast, the centrally controlled fleet typically has vehicles move to the nearest plant needing a vehicle. The deadhead miles for the centrally controlled fleet are approximately 35 to 40 percent of the loaded miles. Merging the two fleets would result in a reduction of empty miles and a reduction in the overall size of the division's fleet.
But by using a modeling tool, management can objectively weigh the potential savings against the total cost of the project. As noted in Figure 3, the potential savings from our manufacturer's project are substantial, with savings that could run into the millions of dollars. The conclusion was that the project could deliver tremendous value for the company.
Ultimately, however, the value of this project will be measured by how much of the potential savings are actually realized. To demonstrate the value of the project as quickly as possible and maintain senior management's support, Recreational Gear decided to begin its network redesign not with the scenario that promised to deliver the biggest savings but with the one that would be quickest to implement.
The implementation timeline often depends on the amount of control and discipline the company (or business division, in this case) has over the distribution network. In addition, some network redesigns require changes in processes and technology in order to become operational. Changes in either of these areas generally involve the extended enterprise. Therefore, it will take considerably more time and effort to implement some scenarios. Given that the distribution network is not static, additional analyses of some scenarios and the environment may need to be conducted to confirm previous results.
Prior to the project's start, Recreational Gear had already begun standardizing the packaging templates that showed distribution facilities how to optimize trailer loads (scenario 1). This scenario also offered one of the largest potential savings of all the analyses performed. Because the decision about cubic utilization is done at the plant, it was relatively easy to implement process changes at this "local" level. For these reasons, the company decided to continue with its efforts. The shipping departments of the various units are currently being trained in using the packaging templates, and there is a concerted effort to integrate loading best practices and experience across the business division. In addition to load optimization, it was also relatively easy to implement the use of the most economical carrier. This involved incorporating another rule in the TMS that specified the selection of the most economical carrier when assigning loads.
Merging the private fleets into the centrally controlled fleet (scenario 5) is also one of the top implementation priorities. Because the fleets are a legacy of earlier acquisitions, implementing this change to the distribution network will require management support and cooperation from the affected business units. At this time, the company is reviewing additional operational considerations to ensure that the decision won't have unintended consequences for other areas of the operation—for example, by creating transportation-capacity problems elsewhere in the network. Once this review is completed—barring any unforeseen issues—a time line will be established for the merger of the various private fleets.
Establishing regional pickup hubs (scenario 3) is relatively easy to accomplish operationally. However, the shifting of deliveries from the end of the period to the beginning of the next period would require both internal and external changes. Although the savings from this change would be substantial, holding shipments for three to seven days would have an impact on the overall order cycle time for some customers. This type of change would require the support and buy-in from customers, sales, and the credit and accounting departments for the business units. For this reason, the implementation of this scenario will be the most intensive and require the most time. Currently, Recreational Gear is discussing the possibility of a pilot project to test the feasibility of this scenario.
Network optimization vs. network redesign
Recreational Gear was thrilled with the potential savings from the project—even more so because it didn't require a redesign of the entire system. It also involved no significant investment and little cooperation with other functional areas within the company.
The company was not ready at this time to commit to a complete network design study. Instead it viewed this project as an opportunity to optimize the operations of an existing network design, focusing only on transportation. It was apparent to all concerned that we were optimizing a sub-system within the supply chain. For those companies that are ready to take it, the next step would be to redesign the total system—the entire supply chain of the company. This is where the real savings will occur.
A redesign of the total system involves a "green field" approach, where all options and alternatives are considered. There are generally no limits placed on the network design as the software program determines the optimal number and location of facilities, the size of each facility, and which products customers will receive from which distribution point.
During the course of the project, the team learned several key lessons. As we stated at the beginning of this article, the network is a dynamic structure that evolves and changes over time. In fact, several organizational changes occurred as we were conducting the network analysis, and changes continued to unfold even after implementation had begun. This dynamic environment makes it difficult to fully measure the success of changing the outbound network design. To compensate, the project team has chosen to focus on quick results as a means to keep the business division engaged in implementation of broader opportunities to improve the distribution network's efficiency.
An additional difficulty was that the data collection and validation phase took more time and effort than expected. The lesson here is that an organization cannot assume that because it has a TMS, the data will be "perfect." In fact, in some cases, filling in the missing data proved to be one of the biggest challenges in building the model. All of this translated into a lot of time in the project phase, which is the least interesting to management. Having the support of management was a critical factor in successfully completing this phase of the project. Spending this time up front, however, ultimately led to a shorter time period for validating the model and enabled us to move to optimization of the baseline in a timely manner. Also noteworthy is that by cleaning the data for this project, we were able to provide an accurate and complete history of operations that would not have been available otherwise. This allowed the company to analyze other aspects of its operations, including carrier efficiency, average length of haul by equipment type, and empty backhaul miles by carrier and equipment type.
Perhaps most importantly, the project reinforced our earlier conclusion that by adopting the systematic and logical modeling approach, companies can reduce the level of implementation risk. For example, most of the project team members had assumed that regional delivery hubs would generate transportation savings. Yet this did not turn out to be the case. Without the model, which provided the opportunity to experiment with this design, the company might have proceeded with a plan to set up regional hubs. Because we were able to gain deeper insight and knowledge about this structure from the model, we now know that it is not worth pursuing at this time.
Would we do it again? The answer from both sides is a resounding yes. Recreational Gear will be able to redesign its network to make efficient use of its transportation budget in a fairly short time period. From the University's perspective, we were able to gain additional experience about network analysis and optimization that will enable us to be more effective teachers in the classroom. It doesn't get any better than a win-win for both sides.