Supply Chain Analysis at Volkswagen of America

The article “Supply Chain Analysis at Volkswagen of America” provides a review about the existing distribution system of Volkswagen of America and also specifies the chances for modifications with significant savings. The post was written by 2 external consultants who were individuals in the research study on how to reengineer the distribution procedure at Volkswagen of America. Among the authors is a consultant from IBM Supply Chain Optimization Practice at Michigan and another is a specialist from Production Modeling Corporation at Michigan as well.

The third author of the article was an actual employee at Volkswagen of America located in Michigan. The real his position is not pointed out in the post.

In 1995, Volkswagen of America appointed a project team to examine the existing automobile distribution procedure and to develop a new enhanced circulation model. The business’s objectives were to improve responsiveness and lower costs from consumer to customer throughout the world. Volkswagen of America, a subsidiary of Volkswagen AG (Germany), imports and disperses Volkswagen and Audi cars in the United States.

The automobiles put together in Mexico or Germany are dispersed to a network of Volkswagen dealers across the United States.

The authors mention that the existing distribution system had actually served the business for many years and had actually been seldom taken a look at. The primary argument is that today system was not designed to maximize profit for the entire system, but only to recognize capital for the original manufacture. That is, the interest was only the advantage of manufacture instead of benefit for Volkswagen and dealerships in the United States together.

The existing circulation system in the United States “has a single dominant type, in which the initial devices produces (OEMs) inventory and sell brand-new lorries to franchised sellers.” So, the dealership was a primary client for a Volkswagen, and the dealership was a provider to the end user. As a result, dealerships and OEMs were just loosely connected within the system developing many drawbacks for both celebrations.

Dealers were encouraged to carry as much inventory as possible, even though realizing that it was threatining for their businesses. Moreover, they were dependent on OEMs because they could restrict supply or appoint competing dealers. This specific distribution logic was developed under assumption that vehicles were configured as a standardized product line- one model in one color. Even after introducing a competitive advantage as a primarily focus for the franchise network, the structure of the distribution system had remained the same. The result of this structure was that dealers were not able to offer the customer an actual required product because of the limited inventory, even though OEMs in the last decade started to offer thousands of model designs.

The subject of this paper is a study about a new configuration that could deal with these limitations and achieve benefits for Volkswagen and dealers together. The goals included to maximize the percentage of customers receiving their first choice of vehicles within 48 hours either from dealer inventory or from Volkswagen. By reducing inventory the goal was to reduce the total system costs for dealers and Volkswagen together, including transportation, financing, and storage costs. One potential solution was to pool vehicles in regional depots so that it would reduce the local dealers’ inventories and in the same way would assure delivery within 48 hours. So, the system would increase first-choice sales, be timely, and reduce inventory costs. However, Volkswagen had no effective way to test this system for feasibility. The team of the study was concerned about integrating such a system with the existing seaport-based distribution centers. The complexity and scope of the system made static analysis of limited value. So, the team of the study identified simulation as a potentially effective way to test the model and various scenarios of implementation.

By the new potential model, the vehicles for sale would first be shipped to five US ports that would act like distribution centers. Then they would transport vehicles by rail or truck to dealers across the United States. The study specifically focused on simulating the flow of the vehicles from plants to dealers. The main idea was to have more distribution centers close to metropolitan markets in order to test the hypothesis that it would increase customer responsiveness and minimize the costs of the total system. The study team tried to look for potential new locations for distribution centers and their opening sequence. The expectations were that by combining dealer and distribution center inventory, the new system would increase the possibility of supplying customer’s first-choice vehicle, and would reduce dealer’s worry of carrying high inventory. The new system would also replace existing expensive truck routes by cheaper rail routes. All of these factors would lead to lower total system costs and greater customer responsiveness.

The system described by authors consists of customer-flow and vehicle-flow cycles. In the customer-flow cycle, customers arrive at dealers and ask for a vehicle with specific features. If such a vehicle is available within dealer’s inventory, it is a strong benefit from the customer- service perspective. If the dealer does not have a required vehicle, the customer can issue a direct factory order for the first-choice vehicle, can agree to buy a second-choice vehicle or just leave the dealership. The customer-flow cycle defines the customer-service level. The way authors measured customer service included first-choice hits by dealers, distribution centers, and through dealer trades, second-choice hits, direct factory orders, and the number of lost customers.

Another system cycle, the vehicle-flow cycle, starts when dealers order vehicles from the distribution centers to replenish their inventories, and distribution centers order from the plant to maintain their pool inventory. The cost in this cycle depends on the mode of transportation and the mileage between two points. The authors brakes the total cost into three components: the cost from plant to processing center, the cost from processing center to distribution center, and the cost from distribution center to market area. In addition, they added inventory holding costs as a finance charges to the total distribution cost. So, no doubt that the number and locations of processing and distribution centers are the major factors affecting customer service and distribution cost. In this scenario, Volkswagen of America considered two types of facilities. The first type facility was smaller in capacity and cheaper, and the second type facilities were larger and the increase in operating expenses was nonlinear, so that they could offer economies of scale.

First of all, the study group developed location scenarios to specify the number and locations of DC’s and processing centers. They also specified the market areas covered by each DC and processing center. In this scenario, totaling of the performance measures required consideration of the “dynamic and stochastic elements in the system.” As the authors explain dynamic elements included inventory control policies and demand seasonality over the year, while customer demand, customer choice and transportation delays represented stochastic elements.

They used a simulation model to consider both elements. The model included the customer and vehicle cycles to calculate performance measures. In addition, some assumptions were made. First, the number of dealers was divided into 52 major demand areas. Second, they defined 50 products, each representing a family of vehicles with similar characteristics. Third, the team analyzed past sales data to calculate demand variation within a year. Volkswagen of America approximated the probabilities for alternative customer actions.

The study team identified 18 potential locations for DC’s and processing centers in the United States, including the current five ports. They also ran a few alternative scenarios based on intuition. However, because of the tremendous number of alternatives, there was a need for a systematic way of generating location scenarios. In order to reduce the number of alternative location scenarios to be evaluated, the team formulated a mixed integer program (MIP) that generated a reasonable number of scenarios. The formulation was customized from the well-known fixed -charge problem. The way MIP worked was it minimized “a cost function that approximated the distribution cost of the actual system ignoring the stochastic and dynamic elements.” The output of the MIP was a location scenario.

The objective function consisted of two factors: total transportation cost (mileage, modes of transportation, and truck load factors) and, fixed and overhead costs of Dc’s and processing centers. Truck load factors referred to the average number of vehicles that a truck carried in a shipment that was maximum 10 vehicles per truck. The truck load was used to determine the number of shipments to a major market. To make it easier to calculate transportation cost, there was made an assumption that each truck trip cost had a fixed factor and a mileage-variable component. The inventory holding cost was ignored in the MIP. Specific constrains assured that the market demand was satisfied, that the capacities were not violated, that the vehicle could be shipped within time window, and that the number of Dc’s and processing centers did not exceeded maximum.

Two major inputs in the MIP were actual sales and truck load factors. The MIP was a front end to the simulation. The team developed a heuristic procedure that iterated between MIP and simulation until both models agreed on a particular location scenario. If the output location scenario had changed, they ran the simulation using the new location policy as input. Even though, the team “could not guarantee convergence in general, this procedure resulted in fairly quick convergence in their computational experiments.” Most of the time they reached a final location scenario between MIP and the simulation in two to three iterations, and the total number of iterations never exceeded six.

The simulation model was implemented using Promodel software. They coded the MIP using AMPL modeling language and used CPLEX as its solver. The team used text files and Excel macros to automate the communication between the MIP and simulation. The existing system with five DC’s and processing centers was a benchmark for subsequent scenarios. The team first devised the best-case scenario, and then they generated a number of interim scenarios that defined a path to follow from the existing to the ideal scenario. In the first scenario, they assumed that all vehicles had to go through processing centers, which was a current case and then to DC’s. The maximum number of DC’s was set to six. The scenario produced the DC which gave the highest profit. From where, they started to increase the number of DC’s fixing only the existing five centers. They tried this scenario until the opening new DC was not profitable anymore. For another scenario, they repeated this analysis under the assumption that vehicles did not need to go through processing centers, but instead, went directly to a DC. Finally, they examined the effects of decreasing the number of processing centers and the effects if the other DC’s had acted as processing centers by simply adding additional constrains in the MIP.

The major findings of their quantitative analysis based on the optimization and simulation results showed that certain modifications made to the distribution structure could result in more customers receiving first-choice vehicles and simultaneously reducing total system cost. For example, the total cost of installing and operating pool facilities were not significant compared to the savings in transportation costs. Volkswagen had opened a number of pilot DC’s to test the implementation of these findings and realized varying degree of success. The finding stated that dealers in the southeast realized less benefit because they were in close proximity to the existing ports of entry, while dealers in midwestern markets experienced improved service and reduced cost. The new strategy was not accepted universally. For example, southeastern dealers saw the reduction in inventory as a potential loss in competitive advantage. In 1998, after introducing a number of new models to the US market, the pilot DC’s were not able to maintain the required stock level, because of the very high demand. So, in 1999, the pilot program was withdrawed.

The optimization tools presented potential opportunities to improve supply-chain by focusing on system level optimization rather than just local. Even though “the pilot test was ended, the learning from the simulation and testing had not been lost.” Most importantly, Volkswagen realized that supply-chain must be viewed from the system level. The existing system focusing on cash-flow could only minimize total system profitability and customer service.

The article “Supply Chain Analysis at Volkswagen of America” is very well written and self- explanatory. It is very easy to read and to understand how model is created and how it works. The advantage is that you do not need to be majoring in supply chain techniques in order to get the main idea. Personally, I really liked the article. I found it very useful and as mentioned above very easy to read and understand. The authors did a great job in portraying the existing supply chain with its disadvantages. They showed good reasons for improvements and necessity for change.