Adding a second production line or a new facility fundamentally breaks a single-line Theory of Constraints implementation. The elegant, localised TOC model that successfully drove throughput on your original floor cannot handle the increased complexity of multi-site operations. As volume scales, the precise drum-buffer-rope synchronisation you perfected begins to generate systemic chaos rather than resolving it.
The failure is structural and entirely predictable. In a single-line environment, identifying the active constraint requires monitoring a linear, physical flow of materials. When you scale operations across a network of facilities, the system's constraint often shifts from a physical machine to a logistical interface, a shared supplier bottleneck, or a specific engineering release. Eliyahu Goldratt's math remains flawless, but the operational environment supplying the variables has changed entirely.
The resulting operational paralysis manifests as a severe scale-up quality crisis. Local plant managers, driven by site-level efficiency targets, aggressively optimise their individual facilities. They flood the internal supply chain with work-in-process that the network's true bottleneck cannot absorb. The customer experiences this localised optimisation as systemic delivery failure, chronic part shortages, and degrading product quality driven by excessive material handling.
How Product Mix Complexity Obscures the True Constraint
Scaling a manufacturing operation almost always introduces a wider, more volatile product mix. A dedicated line producing a single automotive bracket has one easily identifiable physical constraint. When that same site transitions to building ten complex assemblies across multiple customers, the bottleneck begins to shift by the hour. A resource that had ample capacity yesterday becomes the gating factor for total system output today.
The scheduling system rarely keeps pace with this volatility. Material Requirements Planning systems assume a static routing environment, but a complex product mix turns static routings into a liability. As the product profile diversifies, a station that was previously a non-constraint suddenly dictates plant output because of a unique machining requirement on a new part number. If the scheduling team is still anchored to the previous month's capacity model, they will push work into a ghost constraint.
Validating the constraint under high mix complexity requires a different operational discipline. You cannot rely on a historical snapshot of machine utilisation. The scheduling team must actively map the intersection of the daily demand mix against available capacity. If a specific product family routes through a shared heat-treat furnace or a specialised testing station, that resource becomes the network's structural constraint. Running a high-volume, low-complexity part through the system to maximise local OEE actively destroys customer delivery.
Across two decades in automotive and aerospace manufacturing, I have seen plants aggressively scale their output without scaling their constraint detection logic. The most damaging failure mode is treating a temporary, mix-induced bottleneck as a permanent capacity limitation. Organisations throw capital at a machine that looked busy for three weeks, completely missing the systemic constraint sitting at the surface finish line or the final inspection gauge.
The Death of Local Efficiency in Multi-Plant Networks
When a manufacturing footprint expands to multiple sites, the most toxic policy constraint becomes entrenched: the isolated site-level efficiency report. Each plant operates as a profit centre, measured independently on machine utilisation, labour efficiency, and standard hours earned. The corporate dashboard actively incentivises plant managers to build inventory the broader network cannot consume.
This localised optimisation creates massive systemic friction. Facility A meets its monthly OEE targets by running its stamping presses at full capacity. Facility B, which assembles those stamped components, is overwhelmed by the fluctuating inventory levels. The material sits in transit or in staging areas, accruing damage and generating non-conformance reports. The corporate quality team generates dozens of 8D reports investigating handling damage that is actually a symptom of a broken subordination policy.
Traditional cost accounting is the mechanism that enforces this failure. Standard costing models reward facility managers for absorbing overhead through high machine utilisation, regardless of downstream demand. Throughput accounting, which Goldratt advocated precisely to prevent this failure, measures the entire network's throughput against the true constraint. Implementing TOC across a multi-site network requires completely dismantling the standalone financial metrics that actively reward local inventory creation.
Scaling TOC: Site-Level vs Network-Level Optimisation
Site-Level Optimisation
- Plant managers measured on local OEE and labour utilisation
- Cost accounting rewards high machine output regardless of demand
- Facilities push excess WIP into the internal supply chain
- Network experiences systemic congestion and quality degradation
Network-Level TOC
- Subordination enforced across all sites to protect the true constraint
- Throughput accounting measures total system delivery, not local output
- Non-constraint plants deliberately run below capacity to match demand
- Internal supply chain synchronises flow directly to customer requirements
Subordinating the Supply Chain to a Single Point of Failure
The third focusing step demands that every non-constraint resource subordinate itself to the pace of the constraint. In a scaled manufacturing operation, this means entire facilities, logistics networks, and tier-2 suppliers must intentionally slow down. They must produce less than they are capable of producing. This level of cross-functional subordination is where almost all multi-site TOC implementations catastrophically fail.
The behavioural resistance is fierce and immediate. A plant director measured on asset utilisation will not voluntarily idle a production line because a downstream facility at another geographic location is bottlenecked. A procurement director will not reduce component orders when their supplier scorecard rewards on-time delivery to the receiving dock, even if those components feed a non-constraint process. The corporate metrics system actively destroys the subordination required for TOC to function.
When subordination breaks down, the network's constraint is buried in uncoordinated inventory. A surface mount technology line producing cable assemblies pushes forward aggressively, flooding the final assembly area. The assembly constraint chokes on the excess material, lengthening changeover times and destroying batch traceability. Hot lists and emergency expediting become the standard operating procedure, completely overriding any semblance of planned drum-buffer-rope synchronisation.
Sustaining subordination across a multi-site footprint requires structural measurement changes, not interpersonal willpower. Corporate operations must strip local utilisation metrics from the executive compensation structure. The primary KPI for every facility manager must become the throughput of the entire network, measured against the identified systemic constraint. Without this fundamental realignment of incentives, the subordination step is abandoned within weeks.
Constraint Migration in a Dynamic Manufacturing Footprint
If a scaled manufacturing operation successfully identifies and exploits its systemic constraint, the constraint will inevitably move. This is the fifth focusing step, and it is the ultimate test of a mature TOC implementation. In a dynamic, multi-site environment processing a shifting product mix, the constraint can migrate from a machining centre at Plant A to a paint booth at Plant B within a single production cycle.
Most multi-site organisations completely miss this migration. Their operational metrics, daily meetings, and management attention remain locked onto the original constraint. The dashboards built to monitor that first bottleneck become the organisation's structural blind spot. The actual constraint runs completely uncontrolled, while the network continues to subordinate its output to a resource that now has ample excess capacity.
The physical constraint is usually easy to find; the policy constraint is what actually limits your system.
Organisations must treat constraint identification as a continuous, network-wide diagnostic rather than an annual strategic assessment. In a complex aerospace assembly environment producing multiple variants daily, the constraint can shift in hours. Validating the system's true bottleneck requires testing whether reducing capacity at a suspected resource actually drops total system throughput. If the output remains unchanged, the scheduler must immediately look elsewhere.
Scaling Constraint Identification Across Facilities
- 01Map Network DemandAnalyse the intersection of customer volume and product mix against total network capacity.
- 02Isolate the Systemic ConstraintIdentify the single resource whose capacity limits total system throughput across all facilities.
- 03Enforce Network SubordinationDeliberately run non-constraint sites below capacity to match the pace of the true constraint.
- 04Validate Constraint StatusConfirm that removing one unit of capacity from the constraint drops system output.
- 05Repeat DailyRestart the cycle every shift to catch constraint migration before it disrupts customer delivery.
Replacing Consultant Dependency With Embedded Discipline
Scaling a TOC implementation usually begins with a corporate initiative led by external experts. The consultants map the value stream, facilitate the Thinking Processes, identify the systemic constraint, and mandate a subordination plan. The dashboard improves. The consultants depart. Within three to six months, the multi-site implementation is dead, and the network has reverted to local optimisation.
The consultant-driven model creates an illusion of competence transfer. The central team observed the analysis but never owned the daily discipline of constraint validation. When the constraint migrates from a sub-assembly line to a final assembly station, no internal scheduler has the authority or the analytical skill to re-run the five focusing steps. The network experiences this failure as a sudden, unexplained collapse in on-time delivery.
Goldratt's Thinking Processes—the Current Reality Tree, the Evaporating Cloud, the Future Reality Tree—are rigorous analytical tools for untangling complex, multi-site systemic conflicts. In the hands of an untrained team, they devolve into conference-room exercises that justify existing policy constraints rather than challenging them. The Current Reality Tree becomes a complaint board. The Evaporating Cloud is used to validate the standard cost accounting system rather than dismantling it.
Sustaining TOC across a scaled operation requires a permanent operational capability, not a consulting project. The plant managers and network schedulers must be trained to identify constraint migration independently. When subordination breaks down under pressure from a high-volume month, the local leadership must have the discipline to acknowledge the failure and re-align the flow. Corporate must support this by measuring network throughput, not local utilisation.
The physical bottleneck is ultimately the easiest element to identify and, with sufficient capital investment, to elevate. The real constraint in a scaled manufacturing environment is the set of policies, measurements, and financial reporting structures that prevent the organisation from managing that physical bottleneck. Scaling the Theory of Constraints is an organisational design methodology, not a scheduling trick.