Locating your distribution centers Part 1: Re-engineering the distribution center network

Adding more DCs closer to your customers may mean lower transportation costs and higher service levels, but it also means an increase in operating and fixed costs—not to mention the capital cost of a new building. Here’s how to arrive at just the right network at a reasonable cost.

By Maida Napolitano, Contributing Editor -- Logistics Management, 2/1/2008

Why re-engineer the distribution network?
Modeling is the key
Detailing the critical steps
Next steps
Network Study Advice from Our Experts

Editor’s Note: Part 1 of our 2-part series on locating your distribution centers explains how you can re-engineer your entire distribution network by modeling with software to evaluate it under various scenarios to find the best general geographic area. Part 2 will focus on how to nail down the specific site of your new distribution center (DC).

Raise your hand if you happen to be one of those East Coast–based companies that receives product from Asia in Northeast ports and then ships product back across the country to service West Coast customers.

It wasn’t too much of a problem when only 10 percent of your SKUs were manufactured overseas, but you may be feeling the pinch now that 80 percent of your items are sporting “Made in China” labels. Let’s face it: It just may be time to restructure your entire distribution center (DC) network.

Determining the correct number and locations of distribution centers to ensure the effective delivery of products from the source of creation to the point of demand is one of the most challenging strategic planning problems facing logistics managers. Adding more distribution centers closer to your customers may mean lower transportation costs and higher service levels, but it also means an increase in operating costs and the fixed costs for maintaining multiple facilities—not to mention the capital cost of a new building investment.

The goal for optimizing a distribution network is deceptively simple: Find the best balance of costs, customer service, and performance. How do we achieve that goal? Well, that’s were it becomes complicated.

For the past two decades, Jack Kuchta, vice president of Gross & Associates (G&A), a Woodbridge, N.J.-based logistics consulting firm, along with Bill Elenbark, a senior engineering consultant with G&A, have been re-engineering distribution networks for a diverse global clientele. Using the latest location modeling software, they have recreated distribution networks on the computer, experimented with different scenarios, and recommended the most appropriate networks to meet their clients’ corporate goals.

In Part 1 of this two-part series, Kuchta and Elenbark, along with two of their top network modeling clients, share the process they go through when re-engineering distribution networks to improve product flow through global supply chains. Here’s what they’ve learned.

Why re-engineer the distribution network?

The most common driver for distribution center (DC) network re-engineering, says Kuchta, has been the realization that freight costs have a significant dollar impact on the bottom line. “With today’s rising energy costs, reducing transportation costs across the network has certainly become more pervasive and cannot be ignored,” he says. A network study may simply stem from operational growth resulting in a need for an additional warehouse in a different geographic area.

“Opening more stores in the West Coast, for example, may trigger a need to add a West Coast distribution center (DC),” adds Elenbark. Other reasons for studying your network may include mergers or acquisitions resulting in overlapping or excess warehouse space, new company policies and buying incentives forcing a shift in freight modes, or a need to reduce warehouse costs by investigating other areas with more affordable labor pools.

John Goione, who oversees the distribution of service parts and accessories to BMW’s retail group of car dealers and showrooms, found that transportation costs were just part of his bigger concern. “In 1999, we had two distribution centers, one in the East Coast and another in the West Coast,” recalls Goione. “But continued growth volumes, a hypercompetitive premium luxury automotive market, and rising transportation costs all compelled us to revisit our network.”

Steve Burley, logistics manager for Crosman Corp., a manufacturer and distributor of airguns and other sporting goods, had a different catalyst for his company’s distribution center (DC) network study. “Our Bloomfield, N.Y., facility was sufficient when we were simply an airgun manufacturer. However, when we began importing Soft Air and accessories, we quickly outgrew it.” Burley and his team could have simply leased more space, “but it would not have been an informed decision,” adds Burley. “The network study became a tool for us to analyze our data and to explore multiple facility scenarios based on our customer base—and make the proper, informed decision.”

Modeling is the key

Modeling is the process of using historical data to develop a graphical and/or mathematical representation of an actual object or system with the goal of saving costs and minimizing risks through experimenting with the model—not the real system.

In the world of distribution networks and global supply chains, there is much to consider when it comes to successful modeling. Inbound and outbound transportation costs, warehouse fixed and variable costs, freight modes, product lead times, and service zone limits are only a few of the data points required. Multiplied across three or four facilities, accurate analysis becomes impossible unless a computer model of the entire network is created.

Computer modeling certainly can’t happen overnight. “The modeling process takes about four to six weeks and generally costs between $25,000 to $100,000 depending on complexity and the scenarios that a client wants to test,” explains Kuchta. He warns, however, that the cost of actually implementing the revamped network is what’s impossible to define. “The actual cost of relocation, employment packages, and systems changes, can vary widely for each company,” he explains.

Still, the benefits are undeniable. “Reengineering the network can generally save a company upwards of 10 percent of its distribution network costs annually—which could translate into millions of dollars,” says Elenbark. The optimized distribution center (DC) network may also have lower annual operating costs, increased service levels to customers, optimized allocation of inventory and resources between distribution centers, and a theoretically-tested game plan that can handle shipments of future growth levels.

Detailing the critical steps

The creation of a network model is often an iterative process where the initial goal is the validation of the network model before any alternatives can be tested. The following approach outlines the critical steps towards achieving the best network for your distribution needs.

Define the problem and establish the scope: What is your primary objective in studying your network? Is it to minimize costs and/or to find the network that will provide the best customer service? According to Goione, “Our goal was to get us closer to our customers. The new network needed to allow for high service levels demanded by our intense market segment competition.” The scope must also be established. After thinking it through, you may find that only a portion of the supply chain or a certain product group will need to be included.
Collect and analyze the data: “We basically start with a shipment file for six months to a year of customer orders by product group from each existing warehouse,” explains Elenbark. “We also need details such as quantity shipped, weight and/or cube, along with the inbound source for each product group.” Goione recalls that the data definition, collection, and organization took the longest time in BMW’s study. Crosman’s Burley adds that although most of their data was readily available, “providing it in a readable format took a few days,” he says.
Build, run, and validate the model: Once all the data is collected, Elenbark uses a location modeling software that utilizes three to five-digit zip codes to spatially locate suppliers, distribution center (DC)s, and end-destination customers along with actual freight costs to model inbound and outbound transportation costs. To ensure the accuracy of the model, the current network is first modeled and then validated against actual shipments and costs. Elenbark adds, “We will spend several iterations to tweak the model until the results are within 2 percent of actual shipment costs.” BMW’s Goione recalls, “It came very close to reality.”
Develop alternatives: In generating alternatives, the basic elements comprising a validated model are changed to test their impact on total costs. Alternatives may include:
- adding or dropping a distribution center (DC);
- changing the location of one distribution center (DC) while fixing another;
- changing the mode of shipping and associated freight costs by location or by product group;
- changing the objective to “best” customer service instead of minimizing costs.

Burley adds that one alternative they tested was when Wal-Mart offered the option of paying the freight versus Crosman picking up the freight charges. Certain constraints may also have to be imposed on the model. Goione points out that they needed their distribution center (DC)s to be near major carrier hubs. All these were tested on the model and the results tabulated.

Compare alternatives: After each iteration, the following results are usually presented:
- inbound costs from each supply point to each distribution center (DC);
- outbound costs from each warehouse to each customer or customer distribution center (DC);
- a determination of which customer is serviced by a particular distribution center (DC);
- cost of inventory;
- suggested product allocation.

For each alternative, costs are tabulated and compared. If key objectives of the study are not met, additional alternatives may be pursued.

Analyze peripheral issues: The mathematically-optimal network may not be the most practical solution. Management may point out other issues that may influence the solution. Burley recalls how the model was leaning towards opening a new distribution center (DC) in Cincinnati, Ohio. “However, being as the model brought us that close to the New York area, we decided that Rochester, N.Y., would be a better location, as we have easy access to Canada and we are only a day’s travel to the port for exporting.”
Develop the final network design: The new network solution is often a time-phased process with each major change implemented in stages. “BMW’s network solution is still evolving and we haven’t stopped since,” says Goione. “Our new network consists of three pairs of two distribution centers.

The two distribution center (DC)s in each pair are split between North and South with the final location being predicated on sales volume penetration and the desire for proximity to our customer base,” he adds. For example, BMW’s East Coast cluster is comprised of a N.Y. Metro Area distribution center (DC) and a northern Florida distribution center (DC). The West Coast cluster includes a distribution center (DC) in Stockton, Calif., and one in Ontario, Calif.”

Next steps

If the network solution calls for the addition of a new distribution center (DC) in a new location, the next steps revolve around identifying the top candidate communities and sites within the target areas and working with realtors to do detailed field investigations of the candidates.

Network Study Advice from Our Experts

Bill Elenbark, Gross & Associates:

On exploring alternatives: “Exhaust all alternatives in the computer model rather than in real life. You can conduct product group specific model tests allocating products to specific distribution center (DC)s. Test the effects of shipping directly to customer’s stores instead of their distribution center (DC)s. Test shifting from daily parcel deliveries to weekly LTL shipments.”

On the possible pitfalls of a network study: “One major pitfall is not considering all of the factors that go into a network model. For instance, focusing only on outbound transportation costs and not considering inbound source locations and the costs to get the product to your distribution center (DC), nor the facility costs associated with storing and shipping your product.”

John Goione, BMW:

On the modeling process: “Do the analysis. Spend the time to get the data right and model the heck out of the data. The what-if scenarios are crucial to collecting the right information in order to make good decisions.”

On how to deal internally within your company: “Align with your financial folks closely. You need to determine whether you should invest heavily in brick and mortar or configure for a more adaptive supply chain.

Steve Burley, Crosman:

On the modeling process: “You need to be specific as to what your needs are. It’s much easier to make decisions when you know that you covered all the bases. Modeling the network gave us that security.”