Supply Chain Network Design: Warehouse Location, Make vs Buy & Transportation Cost Modeling

What Is Supply Chain Network Design?

Supply chain network design answers the fundamental strategic question: what is the optimal physical configuration of our supply chain to serve our customers at the lowest total cost and highest service level?

Network design decisions include:

• Facility footprint: How many facilities (factories, warehouses, distribution centres, cross-docks) are needed? Where should they be located?
• Facility roles: Should each facility be a full-range distribution centre, a regional hub, or a specialist single-category operation?
• Product flow: Which facilities serve which customers? Which manufacturing plants supply which DCs?
• Make vs buy: What should be produced internally vs outsourced to contract manufacturers or 3PLs?
• Transportation modes: Which lanes should use TL vs LTL vs rail vs parcel, and at what volumes does mode switching make economic sense?

Network design decisions are capital-intensive and long-horizon. Warehouses are typically leased for 5–10 years; manufacturing investments for 10–20 years. Getting the design right can deliver a 10–25% reduction in total logistics and supply chain costs — and getting it wrong creates a structural cost disadvantage that is very slow to unwind.

When to Redesign Your Network

Network design should be triggered by any significant change in the demand or supply landscape. Common triggers include:

Best practice is to conduct a full network design review every 3–5 years even without specific triggers, and a quick optimization review annually as part of the IBP capital planning process.

Warehouse Location Models

Locating warehouses and distribution centres is fundamentally an optimization problem: minimize total cost (facility cost + transportation cost) subject to service level constraints (maximum delivery time or distance to customers).

The Core Trade-off: Centralization vs Decentralization

• Fewer, centralized DCs: Lower fixed facility cost (fewer facilities to build and operate); lower inventory requirement (risk pooling reduces safety stock — see the Safety Stock Guide); but higher outbound transportation cost and longer delivery distances
• More, regional / decentralized DCs: Higher fixed facility cost; higher inventory (safety stock cannot be pooled); but lower outbound transportation cost and faster delivery to customers

The optimal number of facilities is where the total of fixed costs, inventory carrying costs, and transportation costs is minimized. As transportation costs rise (fuel) or delivery speed requirements increase (e-commerce), the optimal balance shifts toward more, smaller, regional facilities.

The Center of Gravity Model

The center of gravity (CoG) model is the simplest warehouse location tool. It finds the location that minimizes total transportation distance (or cost) by computing a weighted average of all customer locations, using volume (or revenue, or weight) as the weighting factor.

Formula

Where (x_i, y_i) are the coordinates of customer location i and V_i is the demand volume (tons, orders, pallets) at that location. (X*, Y*) is the optimal facility location.

Worked Example

A distributor serves five customer clusters:

Optimal X* = 14,000 / 2,600 = 5.38

Optimal Y* = 14,000 / 2,600 = 5.38

The center of gravity suggests a warehouse at approximately grid position (5.4, 5.4). This starting point would then be reviewed against real candidate locations for land availability, labor market, road infrastructure, and lease cost.

Limitations of the CoG Model

• Assumes transportation cost is proportional to straight-line distance — real-world road constraints and zone-based carrier pricing are not captured
• Only optimizes a single facility location — does not solve the multi-facility network problem
• Does not account for fixed facility costs, which often drive the optimal number of facilities more than transportation costs do
• Best used as a first-cut screening tool, not a definitive location decision

Advanced Location Models

P-Median Model

The P-median model finds the optimal locations for exactly P facilities (where P is specified in advance) that minimize the total demand-weighted distance between customers and their assigned facility. Unlike CoG, it simultaneously solves facility locations and customer-to-facility assignments.

• Useful when the number of facilities is constrained by policy or capital budget
• Solved as an integer programming problem; practical implementations use heuristics or specialized software for large instances

Set Covering Model

The set covering model finds the minimum number of facilities required to ensure that every customer location is within a maximum specified service distance (or time). It answers: "How many DCs do we need to guarantee same-day or next-day delivery to all customers?"

• Particularly relevant for B2C and time-critical supply chains
• Can be modified to the maximum covering model (maximize the demand covered by a fixed number of facilities) when full coverage is not required

Mixed Integer Linear Programming (MILP) Network Optimization

Full network optimization models use MILP to minimize total network cost (fixed + variable + transport) subject to constraints including capacity, service lead time, demand fulfillment, and facility count limits.

Where f_j is the fixed cost of opening facility j; y_j is the binary open/close decision; c_ij is the unit cost of serving demand i from facility j; x_ij is the flow from j to i; d_i is demand at i; and C_j is capacity at j.

MILP solutions require dedicated optimization software (Python/PuLP, Gurobi, CPLEX, or commercial supply chain network design tools). For organizations with hundreds of customer zones and dozens of candidate facilities, this is the industry-standard approach.

Model Comparison

Make vs Buy Analysis

The make vs buy decision determines whether an activity — manufacturing a component, running a warehouse, providing a logistics service — should be performed internally or outsourced to a third party. It is one of the most consequential supply chain strategy decisions, affecting cost structure, flexibility, risk, and competitive positioning.

Total Cost of Ownership (TCO) Framework

The analytical foundation of make vs buy is Total Cost of Ownership — the all-in cost comparison between internal production and external supply, including costs that are often underestimated or ignored in a simple unit-price comparison.

The Make vs Buy Decision Framework

TCO provides the economic answer. Strategic factors determine whether the economic answer is the right answer. Use the following two-step framework:

Step 1: TCO Analysis
Calculate all-in cost per unit for make and buy scenarios over a 3–5 year horizon. Include volume sensitivity analysis (what happens to unit cost at −20% and +20% volume?).

Step 2: Strategic Filter
Apply the following strategic criteria regardless of the TCO verdict:

Make vs Buy Worked Example

A manufacturer evaluates whether to continue making Sub-assembly X internally or outsource to a contract manufacturer (CM):

On a pure TCO basis, buying ($26.80) is marginally cheaper than making ($27.90). However, Sub-assembly X contains a proprietary circuit design that is central to product differentiation. Strategic filter outcome: Make is preferred despite the small TCO disadvantage, because the IP risk of outsourcing outweighs the $1.10/unit cost premium.

Transportation Cost Modeling

Transportation is typically the largest single cost in a logistics network — often 40–60% of total logistics spend. Accurate transportation cost modeling is essential for network design, as the trade-off between facility cost and transportation cost drives most location decisions.

Transportation Cost Structure

Mode Cost Comparison

Building a Transportation Cost Model for Network Design

For network design purposes, transportation costs must be modeled as a function of origin-destination lanes — not estimated as a flat percentage. The steps are:

1. Map all lanes: For each combination of supply point (plant, supplier, DC) and demand point (customer zone), calculate the distance and/or transit time
2. Apply mode-specific rate tables: Use carrier rate cards or freight audit data. For FTL, apply a $/mile or $/km rate plus minimum shipment charges. For LTL, apply zone-based rate tables with weight breaks. For parcel, apply zone + weight matrices
3. Estimate shipment frequency and load factors: Transportation cost per unit depends heavily on how well loads are consolidated. A DC that ships TL direct to major customers is far cheaper than one that ships LTL to many small customers
4. Model mode shift opportunities: At certain volume thresholds, shifting from LTL to TL, or from TL to rail, can reduce cost significantly. Model these breaks explicitly
5. Include inbound costs: The facility that minimizes outbound cost may be far from suppliers, incurring high inbound freight. Total transport cost = inbound + outbound

Total Network Cost Framework

A complete network cost model combines all cost elements to compare alternative network configurations. For each candidate network design (e.g., 1 national DC vs 3 regional DCs vs 5 local DCs), calculate:

Example: 1 DC vs 3 DC Comparison

In this example, the 1 DC network is $2.9M/year cheaper, but the 3 DC network delivers an average of 1 day faster. Whether the 1-day service improvement justifies the $2.9M premium is a business decision — but it must be made explicitly, not by default.

Frequently Asked Questions

Related Resources