Distributed manufacturing: many small plants instead of one large one

One process, replicated, or several processes drifting apart

The commitment is to a master process definition every site follows: same specification, same critical parameters, same test methods, same release rules. Sites still differ in equipment vintage, water quality, ambient conditions and operator experience, so equivalence has to be demonstrated rather than assumed — usually through cross-site sample comparison and a periodic round-robin on test methods. Drift is quiet and cumulative. A local supervisor adjusts a parameter to solve a local problem, nobody records it, and eventually the network is making two products under one specification. Exchanging people between sites catches more of this than any document review, because the differences that matter are usually habits rather than written instructions.

What makes a product worth making near its customer

Distribution economics decide most of it. Bulky items, products that are largely water or air, short-shelf-life goods and anything customised late all lose more to freight and lead time than they gain from concentration. Local content requirements in regulation or contracts push the same way, as do customers wanting short-notice supply. Products with high value density, heavy process capital or demanding cleanliness requirements usually belong in one place, because duplicating those conditions costs more than shipping the output. The honest test compares delivered cost per unit from one large plant against the same figure from several smaller ones, with duplicated overhead counted properly rather than assumed away.

Capital repeated at every node

Each site needs its own core equipment, its own utilities and its own compliance footprint, so total capital rises while utilisation per site falls. That sets a minimum viable size below which a node cannot carry its overhead — a fixed-cost recovery question rather than a production capability one. Expansion happens by adding nodes, attractive because each increment is small and fundable from operations, unattractive because the network's average efficiency never improves with scale the way a single large plant's does. Equipment standardisation across nodes recovers part of that loss, since common machines mean shared spares, transferable operators and one maintenance contract covering the whole network.

Stock falls in transit and rises in total

Holding inventory at several points requires more of it in aggregate: each node needs a buffer against its own demand variation, and the pooling benefit of one stock location is lost. What improves is responsiveness and exposure — less product sitting in transit, and no single warehouse whose loss halts everything. Upstream, suppliers must deliver to multiple addresses in smaller drops, raising their cost to serve and weakening your position on freight terms. Some networks answer that with a consolidation point that breaks bulk for the nodes. Visibility across the network is what keeps the aggregate manageable, letting one node's surplus answer another's shortage before either turns into a purchase order.

Running a network from one system

Recipes, specifications, drawings and revisions have to reach every site at once and demonstrably, which makes controlled document distribution and multi-site master data the backbone of the model rather than an administrative detail. Planning must allocate volume between plants, respect what each is qualified to make, and handle transfers between them. Purchasing lives with a standing tension: consolidated agreements preserve leverage while local buying preserves the responsiveness that justified going local. Recurring failures are thin management attention at small sites, certification costs multiplied across the network, and skills nobody can maintain where one person covers each job.

Frequently asked questions

How do we prove that two sites make the same product?

Treat it as a transfer exercise even when both plants are yours. Compare output from each site against the same specification using the same methods, run samples through a common laboratory, and check the parameters that actually drive the property in question rather than finished attributes alone. Repeat it periodically, because equivalence established at commissioning decays as equipment ages and people change. Document what was compared, so a customer or auditor asking gets an answer rather than an assurance.

Where should quality and process engineering sit in a plant network?

Central ownership of the process definition, local ownership of execution. A small central group holds the specifications, approves changes and runs the comparison programme; each site keeps enough capability to investigate a deviation and act without waiting. The arrangement that fails is fully devolved engineering, which produces local optimisation and divergent processes, followed closely by fully centralised engineering, which sits too far away to fix anything quickly.

Does a spread-out network really reduce supply risk?

It reduces the risk of losing everything at once, provided the sites are genuinely independent. Where all of them draw on the same supplier, the same utility grid or the same specialist maintenance contractor, that correlation defeats the geography. Real resilience requires at least one other site qualified and physically able to make each product, meaning spare capacity held, tooling available, and the transfer exercised occasionally rather than assumed to work.