Distribution Network Design: How to Optimize Warehouse Locations, Number of DCs, and the Cost-Service Trade-off Across Your Network

The cost components of a distribution network

Distribution network total cost has five components, each of which responds differently to changes in the number and location of DCs.

Outbound transportation cost is the cost of moving product from DCs to customers. It decreases as the number of DCs increases, because DCs are closer to customers and average shipment distances are shorter. This is the cost most companies focus on, and often the justification given for adding a new DC. But it is only one of five components.

Inbound transportation cost is the cost of moving product from suppliers into the DC network. It typically increases as the number of DCs increases, because inbound volume is split across more locations, reducing consolidation opportunities and increasing per-unit inbound freight cost.

Inventory carrying cost increases significantly as the number of DCs increases. Inventory must be held at every DC to maintain service levels. The square root law of inventory (see the multi-echelon inventory post) predicts that safety stock increases with the square root of the number of locations: two DCs require approximately 1.4 times the safety stock of one DC serving the same demand. The carrying cost of this additional inventory, capital tied up, warehouse space, obsolescence risk, is often larger than the outbound transportation savings from adding a DC.

Facility operating cost includes rent or ownership cost, labor, utilities, and management overhead for each DC. It increases linearly with the number of DCs. A single large DC benefits from economies of scale in equipment utilization, automation amortization, and management span.

Supply chain complexity cost is harder to quantify but real: more DCs means more inventory balancing decisions, more carrier relationships, more compliance requirements, more IT integration points, and more management attention. Complexity cost is typically underestimated in network design analyses.

The number-of-DCs decision

The total cost curve as a function of number of DCs is U-shaped: too few DCs produces high outbound transportation cost; too many DCs produces high inventory, facility, and complexity cost. The minimum of the curve, the optimal number of DCs, depends on the geographic distribution of the customer base, the density of demand, the weight and value of the products being distributed, and the required service level (same-day, next-day, two-day, etc.).

As a rough rule: networks serving dense, geographically concentrated demand (a city, a region) tend toward fewer, larger facilities. Networks serving geographically dispersed demand (a national or international market) with tight delivery windows tend toward more, smaller regional facilities. The shift to e-commerce and customer expectations for next-day or same-day delivery has pushed many networks toward more DCs and regional fulfillment centers, because outbound last-mile cost and speed are heavily determined by proximity to the customer.

DC location selection: the gravity model and optimization

Once the number of DCs is determined, the location question is: where should each DC be? The simplest approach is the center-of-gravity model, which finds the location that minimizes weighted transportation distance to customers (weighted by demand volume). The center-of-gravity point is calculated as the demand-weighted average of customer latitudes and longitudes. It identifies the transportation-minimizing location but ignores facility cost differences, labor availability, transportation infrastructure, and tax environment, all of which matter in practice.

More rigorous network optimization uses mixed-integer programming (MIP) models that simultaneously optimize DC locations, customer assignments, and inventory positions subject to cost and service constraints. These models are standard in supply chain planning software and can evaluate thousands of location combinations in minutes. The inputs are customer demand data (volume, location, service requirements), potential facility sites (cost, capacity), transportation rates, and inventory parameters.

In practice, network design is as much an art as a science: the optimization model identifies the mathematically optimal network, but real location decisions must account for factors the model cannot capture, proximity to key customers, labor market quality, existing supplier relationships, local regulations, and management preference for certain geographies. The model output is a starting point for structured analysis, not a final answer.

Hub-and-spoke versus direct shipping

Hub-and-spoke networks route product through a central hub (a cross-dock or consolidation center) before distributing to regional spokes (DCs or customers). Direct shipping routes product from the source directly to the destination, bypassing intermediate nodes. The trade-off is consolidation versus speed and simplicity.

Hub-and-spoke networks consolidate smaller shipments into larger ones at the hub, reducing per-unit transportation cost on both the inbound (supplier to hub) and outbound (hub to spoke or customer) legs. They are efficient when demand at any individual destination is too small to fill a direct truckload shipment. The trade-off is additional handling and transit time at the hub, and the vulnerability of the entire network to hub disruption.

Direct shipping is faster and simpler but requires sufficient volume between origin and destination to justify a direct lane. It is most efficient for high-volume flows, full truckloads between a major production facility and a large DC, where hub consolidation adds cost without adding value.

Network design triggers

Distribution network design is not a one-time exercise, it should be reviewed when significant changes make the current network suboptimal. Common triggers include: significant changes in the customer base or geographic distribution of demand (a new market entered, a major customer won or lost), changes in service level requirements (a shift from five-day to two-day delivery expectation), significant changes in transportation costs or carrier strategy, changes in the supplier base or sourcing geography (nearshoring, supplier consolidation), facility lease expirations, and mergers or acquisitions that produce redundant network nodes.

Common mistakes

• Optimizing transportation cost in isolation. Outbound transportation cost is the most visible component of network cost, but adding DCs increases inventory, facility, and complexity costs. Total cost optimization often produces a different answer than transportation-only optimization.
• Using outdated demand data. Network design optimization is only as good as the demand data it uses. Customer locations, volumes, and service requirements change. A network designed on three-year-old data may be optimized for a customer base that no longer exists.
• Ignoring the square root law of inventory. The inventory cost of adding a DC is frequently underestimated. Every new location requires its own safety stock, and the aggregate safety stock across multiple locations is higher than the safety stock needed in a consolidated network.
• Treating network design as a one-time project. A distribution network designed five years ago may be well-adapted to the demand geography, service requirements, and cost structure of five years ago, and poorly adapted to today’s. Regular network reviews (every three to five years, or when major triggers occur) are part of mature supply chain practice.

Frequently asked questions

What is distribution network design?

Distribution network design is the strategic decision about how many distribution centers to operate, where to locate them, which customers each DC serves, and how inventory flows through the network from suppliers to customers. It is a high-leverage supply chain decision because the network structure determines the cost floor and service ceiling: the location and number of DCs sets the baseline transportation and inventory costs that operations then executes against. The core trade-off is cost versus service: more DCs reduce delivery lead time and outbound transportation cost, but increase inventory carrying cost, facility cost, and supply chain complexity. The optimal design minimizes total cost, across all five cost components, at the required service level.

What is the center-of-gravity model in distribution network design?

The center-of-gravity model is a simple analytical tool that finds the geographic location minimizing weighted transportation distance from a DC to its customers. It calculates the demand-weighted average latitude and longitude of the customer base, the “center of mass” of demand. The model is useful as a starting point because it gives an intuitive, data-driven location estimate for a single new DC. Its limitations are that it ignores facility cost differences between locations, labor availability, transportation infrastructure quality, regulatory environment, and tax implications, all of which affect the real cost of a DC at any given location. For networks with multiple DCs and complex constraints, mixed-integer programming optimization is more appropriate.

How does the number of DCs affect total supply chain cost?

The total cost curve as a function of number of DCs is U-shaped. With too few DCs, outbound transportation cost is high because product must travel long distances to reach customers. As the number of DCs increases, outbound transportation cost falls and delivery speed improves. But inventory carrying cost rises (more safety stock spread across more locations, per the square root law), facility operating cost rises (more rent, labor, and overhead), inbound transportation cost rises (smaller, less consolidated inbound shipments), and supply chain complexity increases. The optimal number of DCs is at the minimum of this total cost curve. Adding a DC beyond that point reduces transportation cost but increases total cost, a common and expensive mistake.

What is the square root law and how does it apply to network design?

The square root law of inventory (also called the square root rule for safety stock) states that when inventory is decentralized from one location to N locations serving the same total demand, the total safety stock required increases by a factor of approximately √N. Splitting one DC into two DCs serving equal demand requires approximately 1.41 times the safety stock of the single DC; splitting into four DCs requires approximately twice the safety stock. This inventory cost increase is a direct consequence of losing the risk-pooling benefit of centralized inventory. In network design, the square root law means that every additional DC adds an inventory cost that must be weighed against the transportation savings and service improvement it provides.

When should a company redesign its distribution network?

A distribution network should be redesigned when significant changes make the current structure suboptimal: a major shift in the geographic distribution of customers (new markets, large customer wins or losses), a change in service level requirements (the shift from five-day to two-day delivery expectations that e-commerce accelerated), significant changes in transportation costs or carrier strategy, nearshoring or offshoring of key suppliers, facility lease expirations that provide a natural review point, or mergers and acquisitions that create duplicate network nodes. As a baseline practice, mature supply chain organizations review their distribution network design every three to five years even in the absence of specific triggers, because the cumulative effect of gradual demand shifts can make a previously optimal network substantially suboptimal over time.

What is the difference between a distribution center and a cross-dock?

A distribution center (DC) stores inventory: goods arrive, are put away into storage locations, and are later picked, packed, and shipped to customers. Storage time in a DC may range from days to months. A cross-dock is a transshipment facility: goods arrive from suppliers or upstream nodes, are immediately sorted and consolidated with other inbound shipments, and are dispatched to outbound destinations, typically within hours, with no storage. Cross-docking eliminates put-away and pick labor and reduces storage space requirements, but requires tight inbound and outbound schedule coordination. It is most effective for high-volume, predictable flows where the matching between inbound supply and outbound demand can be planned in advance.