Last updated: 3 November 2026.
Theory of Constraints starts with a deliberately narrow question: what is the one constraint that limits the system’s ability to achieve its goal? The answer may be a machine, approval step, decision rule, scarce skill, market, policy or piece of information. Improving everything at once can waste effort. Improving the constraint can change the performance of the whole system.
- System throughput is governed by its limiting constraint, not by the average performance of every part.
- Local efficiency can make the system worse by creating work in progress before the bottleneck.
- The constraint moves, so improvement is a repeating cycle rather than a one-time project.
Why the bottleneck governs the whole
Imagine a process with five steps. Four can each process 200 items per day, but one can process only 100. The system cannot deliver 200 finished items. Extra work accumulates before the slow step, and downstream workers wait.
This seems obvious in a factory, but the same pattern appears in knowledge work. A team may have plenty of engineers but one security review slot. A hospital may have beds but too few discharge decisions. A company may have data but one approval committee that cannot process the volume.
The five focusing steps
- Identify the constraint. Find what most limits throughput, not what is most visible or easiest to improve.
- Exploit the constraint. Use its existing capacity well. Remove avoidable downtime, interruptions and rework.
- Subordinate everything else. Align upstream release and downstream work with the constraint’s pace.
- Elevate the constraint. Add capacity, change the process, outsource, automate or redesign when exploitation is not enough.
- Repeat. Once the constraint moves, return to the first step and avoid allowing inertia to become the new limit.
The sequence protects the organization from jumping directly to expensive expansion. It also makes a key systems point: the performance of a part should be judged by its contribution to the whole goal.
Why local efficiency can hurt throughput
If an upstream team is rewarded for staying busy, it may release work faster than the constraint can process it. Work in progress grows, lead time increases and defects become harder to find. The local utilization number improves while the customer waits longer.
This is closely related to Little’s Law: when work in progress increases without a matching increase in throughput, time in the system grows. A constraint-focused view asks where that inventory is accumulating and why.
Finding the real constraint
Do not assume the bottleneck is the slowest machine. Look for queues, starvation, repeated expediting, approval delays, scarce expertise and decisions that block many downstream actions. A constraint can be a rule that prevents capacity from being used.
Measure throughput, work in progress and lead time together. A team that completes many tasks may still be constraining the organization if its work is not what the customer is waiting for.
Common mistakes
- Improving a non-constraint because its data is easy to access.
- Keeping the constraint busy with low-value work.
- Allowing every department to optimize its own utilization.
- Forgetting to repeat the analysis after the constraint moves.
- Treating the constraint as a person rather than a property of the system.
A worked example: a service approval queue
Suppose an organization promises a two-week decision but has one specialist approval team. New requests arrive faster than the team can process them. Other departments improve their own productivity and send more complete-looking applications, but the queue grows. Leaders respond by asking everyone to work faster, which increases rework and makes the constraint slower.
A Theory of Constraints analysis identifies the approval team as the current constraint. Exploiting it may mean removing avoidable meetings, improving the information supplied at intake and reserving capacity for decisions that actually affect throughput. Subordinating the rest of the organization may mean limiting release of new work until the queue is within a safe range.
Only after these steps should the organization add capacity or redesign the approval rule. The intervention is judged by completed decisions and lead time, not by how busy every department appears.
Throughput, inventory and operating expense
TOC distinguishes the rate at which the system generates useful output from the amount of work trapped inside it. A process can be highly utilized and still have poor throughput if it produces work that waits, is reworked or is not needed by the customer.
This connects to Little’s Law. When work in progress rises while throughput stays constant, time in the system rises. The constraint often reveals itself through the queue it creates, but the correct response is not always to clear the queue faster. It may be to stop feeding it.
Constraints in projects and knowledge work
In a project, the constraint may be a scarce expert, a decision authority or an integration test environment. In software, it may be code review, release approval or reliable test data. In a policy team, it may be the small group that can reconcile competing requirements.
Knowledge work makes constraints harder to see because work is less uniform. Measure the flow of decisions, usable deliverables or customer outcomes rather than activity alone. A calendar full of meetings may indicate demand on a constraint, not productivity.
Drum-buffer-rope as a coordination idea
The drum is the pace set by the constraint. The buffer protects it from starvation. The rope controls the release of new work so that upstream activity does not overwhelm the buffer. The terms are useful even outside manufacturing because they describe synchronization.
Without a release rule, every team can create more work than the system can finish. The resulting inventory hides the constraint until lead times become unacceptable. A visible buffer and a clear release signal help the organization trade local busyness for predictable flow.
When the constraint moves
Improvement changes the system. Once the approval team can process more requests, a review board, data source or downstream implementation team may become the new constraint. This is not evidence that the original improvement failed. It is evidence that the limiting factor moved.
The danger is stopping after the first success. Repeat the scan, update the measures and ask whether the new constraint is governed by a rule that can be changed before adding more resources.
Limits and cautions
TOC is a focusing discipline, not a claim that all problems have one simple cause. Multiple constraints can interact. Demand can be uncertain. A local increase in throughput can create environmental, quality or human costs that the narrow measure misses.
Define the system goal broadly enough to include safety, quality and sustainability. A bottleneck that is removed by exhausting people or degrading the product has not produced a durable improvement.
Managerial questions
- What is the system trying to produce?
- Where does work wait longest?
- Which resource is protected from interruption?
- What work should not be released yet?
- What will become the constraint after this improvement?
Constraint management is a learning loop
The five focusing steps should be repeated because the system changes after every meaningful improvement. A new capacity investment can reveal a policy constraint. A policy change can reveal a skills constraint. A demand shift can make the market the constraint. Treating the first bottleneck as permanent causes the organization to keep solving yesterday’s problem.
Use a simple review cadence. Track throughput, lead time, work in progress and the current constraint. At each review, ask whether the constraint is being protected, whether upstream release is aligned and what side effect the improvement created.
Do not confuse scarcity with constraint
A scarce resource is not necessarily the governing constraint. A resource can be scarce but irrelevant to current throughput. Conversely, an abundant resource can become a constraint if a rule, handoff or decision process prevents it from being used. Measure the flow of the system before declaring the bottleneck.
Use the constraint to improve the whole
The constraint is a point of leverage, but it is also a point of vulnerability. Protecting it means reducing interruptions, clarifying inputs and making failures visible. It does not mean treating the people or equipment at the constraint as expendable. Sustainable throughput includes quality, safety and the capacity to continue.
A good constraint review therefore tracks both output and strain. If throughput rises while errors, exhaustion or rework rise faster, the system has shifted cost rather than improved. The next improvement should address that side effect before the apparent gain becomes a new failure loop.
Constraint thinking is strongest when it remains connected to the actual purpose of the system.
Protecting the goal
The goal of a system is broader than its easiest metric. Throughput without quality creates rework. Speed without safety creates incidents. Utilization without recovery creates burnout. Before exploiting a constraint, state the conditions that make the output useful and sustainable.
This prevents constraint management from becoming another local optimization scheme. The constraint is a means to improve the whole system, not a reason to maximize one resource at any cost.
Measures that prevent local optimization
Pair throughput with lead time, quality and strain. If the constraint produces more output but creates rework or exhaustion, the system has not improved sustainably. The measure of success is useful flow through the whole system.
When to stop exploiting the constraint
Exploitation has limits. Once avoidable loss is removed, the next step may require investment or redesign. Continuing to squeeze the same constraint can create unsafe workarounds and hide the fact that the system needs a different architecture.
Keep the goal visible
Constraint work is successful when the customer or public outcome improves, not simply when one resource is busier. Keep the goal visible while the constraint moves through the system.
Keep reviewing the constraint after the improvement, because a system that learns will eventually reveal a different limit.
The goal remains useful flow through the whole system.
Frequently asked questions
Does every system have only one constraint?
At a given moment, one constraint usually governs throughput, although several constraints can interact or move quickly. The method focuses attention rather than claiming the system has only one problem.
Is Theory of Constraints only for manufacturing?
No. It applies to projects, healthcare, software, services, supply chains and decision processes wherever a limiting factor governs flow.
What if the constraint is demand?
Then the constraint is outside internal capacity. The response may involve product design, market access, pricing, distribution or creating a new source of demand.
Conclusion
Theory of Constraints is a discipline of focus. Find the factor that limits the whole, protect it from avoidable loss, align the rest of the system to it and then look again. The goal is not to make every part busy. It is to increase useful throughput without creating a larger queue elsewhere.

Discussion