Strategic Supply Chain Network Design: How to Decide Where to Locate

What network design actually decides

Every physical supply chain is, at its core, a network of nodes connected by transport links. The nodes are factories, suppliers' plants, ports, distribution centres (DCs), cross-dock points and retail or customer locations. Network design is the discipline of deciding the structure of that network: how many nodes of each type to operate, where each one is physically located, how large it should be, what function it performs (holding stock, cross-docking, final assembly, returns processing) and — critically — which set of downstream customers or markets each facility is responsible for serving.

This is different from day-to-day operational planning. A warehouse manager decides how much stock to hold this month; a transport planner decides which truck carries which order tomorrow. Network design decides whether that warehouse should exist at all, in that city, at that size. It sets the "playing field" within which all of the tactical decisions — inventory policy, transport mode choice, routing — then have to operate. A well-designed network makes the tactical decisions easier and cheaper to execute well; a poorly designed one forces good operators to compensate, at cost, for a bad structural starting point.

Network design questions typically include: How many distribution centres do we need, and where? Should we hold one national bonded warehouse near the port of entry, or regional DCs closer to end customers? Should a facility hold inventory, or purely cross-dock goods that arrive pre-allocated to onward destinations? Which customers, in which geography, does each DC serve, and does that allocation change by product line or by season? These questions are asked together because they are interdependent — the answer to "how many DCs" changes the answer to "how big should each one be" and "what should each one hold."

The central cost trade-off: fewer big facilities vs more small ones

Almost every network design problem reduces to a single underlying tension, sometimes called the "cost-service trade-off." Consolidating inventory into a smaller number of larger facilities lowers total fixed facility cost (fewer buildings, less duplicated racking and equipment, fewer site management overheads) and lowers total inventory cost, because pooling demand across a wider catchment area reduces the amount of safety stock needed to protect against demand variability (this is a statistical effect — the variability of pooled demand grows more slowly than the number of locations, a phenomenon often summarised as the "square root law" of inventory consolidation). The cost of holding stock to guard against uncertainty falls as you consolidate.

But consolidation has a mirror-image cost on the transport and service side. A single central facility is, on average, further from any given customer than a network of regional facilities would be. That means longer haul distances, higher per-order delivery cost, longer delivery lead times and — for time-sensitive or perishable goods — a real risk that the promised service level simply cannot be met from that distance. Push the network toward more, smaller, more widely distributed facilities and the transport-and-service side of the equation improves sharply: shorter last-mile hauls, faster delivery, better responsiveness to local demand spikes. But you now pay for it in facility cost (rent, staffing, equipment at every site) and in inventory cost (each site needs its own buffer stock, and that stock is no longer pooled across the wider network, so total safety stock held across the whole system rises).

Neither extreme is usually optimal. The job of network design is to find the point on this spectrum — for a given product, market and cost structure — where total network cost (facilities + inventory + inbound transport + outbound transport) is minimised subject to a required service level. That optimum shifts constantly as costs change: fuel price rises push toward consolidation (transport becomes relatively more expensive to run inefficiently); e-commerce demand for next-day delivery pushes toward decentralisation (service requirements tighten); land and construction cost changes shift the facility-cost side of the equation.

The inputs that shape a network design decision

A credible network design exercise pulls together several distinct categories of input data, because the trade-off above cannot be resolved with cost figures alone — it needs a realistic picture of demand and constraints.

• Demand geography. Where are customers actually located, and in what volumes? A network optimised for a demand pattern concentrated in two metros looks very different from one designed for demand spread evenly across a country.
• Transport costs and lead times. The cost per kilometre (or per tonne-kilometre) for inbound and outbound transport, by mode, and the transit time each lane implies — these numbers directly determine how expensive it is to serve distant customers from a given facility location.
• Facility fixed costs. Rent or capital cost, staffing, utilities, material handling equipment and site overheads for each candidate location and size — these determine the true cost of adding another node to the network.
• Required service levels. How fast does an order need to reach the customer, and how reliably? A same-day promise forces a very different, more decentralised network than a five-day promise does.
• Duties, customs and regulatory considerations. For a network that crosses borders, the location of bonded facilities, free trade zones, and the customs duty and landed cost implications of where goods are cleared and stored can be as significant as pure transport economics — a facility sited to defer duty liability or to consolidate customs clearance can materially change total landed cost even before a single outbound delivery is made.
• Currency and country risk. Where a network spans multiple countries, exposure to currency movements on facility leases, imported equipment and inventory carrying cost, plus the political and regulatory stability of each candidate location, are legitimate network design inputs — not just financial footnotes.

In practice, these inputs are modelled quantitatively wherever possible — total landed cost is calculated for a range of candidate network configurations, and the configuration that minimises total cost subject to the service constraint is selected. But no amount of modelling removes judgement from the process: demand forecasts for a five-to-ten-year facility life are inherently uncertain, and a network that is merely "close to optimal" today but flexible enough to adapt is often a better choice than one that is precisely optimal for today's demand pattern but rigid.

Why network design is strategic, not tactical

Network design decisions sit at the top of what is often described as a hierarchy of supply chain decisions, ranked by how expensive and slow they are to reverse. At the bottom of that hierarchy are operational decisions — today's delivery schedule, this week's production sequence — which can be changed daily at essentially zero switching cost. Above that sit tactical decisions, such as how much safety stock to hold or which supplier to order from this quarter, which can typically be revisited every few weeks or months without major disruption. At the top sit strategic decisions — and network design is the archetypal example — which involve long-lived capital commitments (a leased or built facility, installed racking and material handling systems, hired and trained local staff, negotiated long-term transport contracts) that are expensive to unwind.

This asymmetry matters practically. A company can adjust its safety stock policy this month if demand forecasts change; it cannot relocate a distribution centre this month. Because switching costs are high, network design decisions are made infrequently — often reviewed only every three to five years, or triggered by a major event such as entering a new market, a significant shift in the customer base, a change in the cost structure of transport or property, or a merger. Between those reviews, the network is treated as fixed, and all of the tactical and operational decision-making happens within the constraints that structure imposes. This is precisely why getting the analysis right at the design stage matters so much: an under-cooked network decision does not just cost money once, it constrains cost and service performance for years afterward.

A worked example: importing into South Africa

Consider a business that imports finished goods through South African ports and distributes nationally to retail and business customers concentrated in three metro clusters: Gauteng (the largest inland population and economic centre), the greater Durban/KwaZulu-Natal area, and the Western Cape around Cape Town. The company must decide between two broad network structures: one central distribution centre in Gauteng, or three smaller regional DCs — one near each of Durban, Cape Town and Gauteng.

Option A — single central Gauteng DC. Most import volume arrives through the Port of Durban (South Africa's busiest container port) and, for some categories, Cape Town. Under this option, containers are cleared and then moved inland by rail or road — the same sea-vs-air-vs-road mode choice covered in air vs sea vs road: when to use each — to a single Gauteng facility, from which all national deliveries — including back down to KwaZulu-Natal and the Western Cape — are dispatched. This concentrates inventory in one place, minimising total safety stock and facility overhead, and it is operationally simple to manage. But it means goods travelling from Durban to Gauteng and then, for a share of demand, straight back down toward the coast — a long, costly inland leg travelled twice for some of the volume — and it means customers in Durban and Cape Town experience longer delivery lead times than a Gauteng customer does, because every order to those regions has to complete the full inland leg from Gauteng before the final delivery even starts.

Option B — regional DCs near each port/metro. Under this option, imported stock is split and held closer to where it will ultimately be sold: a Durban-area facility serving KwaZulu-Natal (and possibly cross-docking straight off the port with minimal inland movement at all), a Cape Town-area facility serving the Western Cape, and a Gauteng facility serving the inland economic hub. This shortens the last delivery leg for customers in every region, meaningfully improves delivery lead times outside Gauteng, and reduces the double-handling of stock that never needed to travel inland and back. The cost is real, though: three facilities to lease and staff instead of one, and — because demand at each regional DC is no longer pooled with the other two — more total safety stock has to be held across the network to protect the same overall service level, since each region's demand variability now has to be buffered independently rather than being smoothed out across a national pool.

Which option wins depends on the specific numbers: how price-sensitive the product is to inland transport cost, how tight the required delivery lead time is in Durban and Cape Town, how large and variable demand is in each region (larger, steadier regional demand needs relatively less extra safety stock to decentralise), and how significant the facility fixed costs are in each candidate location. A high-value, slow-moving, low-volume product with generous delivery lead times often favours Option A — the inventory pooling benefit dominates, and customers do not notice a few extra days in transit. A high-volume, time-sensitive or low-margin product, where inland transport is a meaningful share of landed cost and same/next-day delivery matters competitively, more often favours Option B, or a hybrid — a Gauteng DC for the bulk of national volume and a smaller cross-dock or satellite facility in Durban and/or Cape Town purely to shorten the last mile for regional customers without carrying full independent safety stock at each site. This hybrid pattern — one strategic inventory-holding node plus lighter-weight, non-stockholding cross-dock points near the ports — is a common practical compromise for South African importers weighing the same fundamental trade-off this article has described throughout.

Putting it together

Network design is not a one-time exercise to be filed away once a facility is built — it is a strategic capability that a business should revisit deliberately as its demand pattern, cost structure and service commitments evolve. The discipline is to be explicit about the trade-off (facility and inventory cost vs transport cost and responsiveness), to gather real data on the inputs that drive it, and to recognise that because the resulting commitments are expensive to reverse, the decision deserves proportionately more rigour than the tactical inventory and transport decisions that get revisited every month. A business that treats network design casually — defaulting to "one big DC" or "a DC in every city" without working through the actual cost and service trade-offs for its specific product and customer base — is very often leaving money, service performance, or both, on the table for years at a time.

Frequently asked questions

What is the difference between network design and inventory management?

Network design decides the structure of the supply chain — how many facilities, where, and what role each plays — and is revisited infrequently because it involves long-lived capital commitments. Inventory management operates within that fixed structure, deciding how much stock to hold and when to reorder at each existing facility, and can be adjusted far more frequently, often weekly or monthly, without any physical or contractual disruption.

Why does consolidating inventory into fewer facilities reduce total safety stock?

When demand from several separate regions is pooled and served from one facility, the variability of the combined demand grows more slowly than the number of regions pooled — high demand in one area is statistically likely to be partly offset by lower demand in another. Because safety stock exists to buffer against demand variability, pooling reduces the total safety stock needed to protect the same overall service level, compared with holding independent buffers at several separate, smaller facilities.

How often should a company review its distribution network?

There is no fixed rule, but a full network review is typically triggered every three to five years, or sooner if a major event changes the underlying assumptions — entering a new geographic market, a significant shift in customer concentration or order profile, a material change in transport or property costs, or a merger or acquisition that brings an overlapping network into the business.

Does network design apply to service businesses, or only to businesses holding physical inventory?

The core logic applies most directly to businesses that store and move physical goods, since the trade-off is explicitly about facility, inventory and transport cost. Service networks (branches, service centres, technician depots) face an analogous but distinct trade-off — coverage and response time against facility and staffing cost — that borrows the same strategic-versus-tactical framing even though inventory pooling is not usually the driving factor.

Should duties and customs clearance really influence where a distribution centre is located?

Yes, for any network that involves cross-border or imported goods. Where and how goods are cleared, and whether a facility operates as a bonded or customs-controlled site, can defer duty liability, simplify clearance, and materially change total landed cost — these effects can be large enough to outweigh a modest difference in pure transport distance between candidate sites, so they belong in the network design analysis alongside transport and facility cost, not as an afterthought.

More concepts in Inventory & Network Design