Bottleneck

A bottleneck is the constraint that limits the throughput of an entire flow. It is the point where work arrives faster than it can leave. In a company, improving steps that are not the bottleneck may make local activity look better, but it does not improve the system result. For SMEs with limited resources, this is a decisive idea: the point is not to improve everything, but to find the point that limits everything else.

In brief

Every operating flow has one point that moves more slowly than the rest. That point, not the average performance of all activities, determines how much the system can produce. In the language of the Theory of Constraints, the bottleneck is the system constraint.

The implication is simple and often counterintuitive: improving a non-constraint does not improve total throughput. It can even make the situation worse by creating more work-in-progress in front of the constraint.

For an SME, bottleneck thinking protects scarce capital and managerial attention. It prevents the company from investing in the most visible pain while the real constraint sits elsewhere.

Operational definition

A bottleneck is the step with the lowest effective capacity relative to the demand it receives. Two observable measures matter.

The first is throughput: how many units of work the step completes in a given period.

The second is the queue that forms before it. The constraint is where the queue grows persistently, not necessarily where people look busiest.

This distinction matters. Downstream from the bottleneck, teams may appear idle or work in bursts because not enough work reaches them. Upstream from the bottleneck, teams may appear extremely busy because they keep producing work that accumulates. Visible pressure and real constraint are not always the same.

Two clarifications keep the term rigorous. First, a system has one active constraint at a time. It can have several weak steps, but one of them limits overall throughput at any given moment. Once that constraint is relieved, the constraint moves somewhere else. Bottleneck work is therefore cyclical, not final.

Second, the constraint is not always a machine, department or technical resource. In SMEs it is often a person with decision authority, a scarce technical competence, or the founder’s attention.

Why it matters for SMEs

The concept matters because investment capacity is limited. If the company invests in the wrong point, it pays twice: money is spent and the real problem remains.

A common pattern is this: the company feels slow, identifies the area that looks most overloaded, hires a person or buys equipment there, and overall lead time does not change. The real constraint was in an administrative, technical or decision step that nobody suspected.

The Theory of Constraints also gives a useful intervention order. First, exploit the existing constraint: remove interruptions, non-essential work and waiting around the constraint. Second, subordinate the rest of the flow to its rhythm: stop launching more work than the constraint can absorb. Only then elevate capacity through investment.

Many SME bottlenecks are reduced in the first two steps, at low cost. Investment is not excluded; it is simply the last step, not the first reflex.

For investors and buyers, the bottleneck answers a central growth question: what breaks first if volume doubles? A management team that can identify the current constraint, the next likely constraint and the cost of elevating them is demonstrating operational capacity more credibly than many growth slides.

Observable signals

Bottlenecks show up in queues. In SMEs those queues have concrete forms: a physical stack of files on a desk, a folder waiting for approval, jobs that remain “in progress” for weeks at the same step, or the recurring phrase “it is with X”, where X is always the same person or department.

A complementary signal appears downstream: teams alternate between waiting and rushing. This feast-or-famine pattern often means work is released in waves by a constraint upstream.

Another signal is temporal. If lead time increases as orders grow while actual cycle time remains stable, the system has reached a saturation point. The queue lengthens at the constraint.

Common mistakes

The first mistake is optimising non-constraints. A faster machine in a step that already waits for work does not improve throughput. Training people to produce faster when their output accumulates at the next step creates local efficiency and system waste.

The second mistake is treating the bottleneck as the culprit. The slow step is not morally guilty. It is often where complexity from the rest of the flow concentrates. Punishing the bottleneck usually makes information worse, because people start hiding the queue.

The third mistake is forgetting that the constraint moves. After the first constraint is relieved, another one becomes limiting. Celebrating and stopping observation means being surprised by the next bottleneck.

The fourth mistake is not recognising human bottlenecks. In growth-stage SMEs, the bottleneck often moves from production to approval, prioritisation or signature. Queues in front of people are less visible than piles of material, but they are often more damaging.

Operational example

A 40-person fabrication company invests 180,000 euros in a faster cutting centre because the cutting area is always full of material waiting. After the investment, job lead time barely changes: from 26 to 25 days.

A later process map shows why. Material accumulated before cutting not because cutting was the real constraint, but because drawings arrived from technical office in waves. Technical office completed six jobs per week while the order book required nine. The apparent queue before cutting was material purchased early “so we do not stop”.

The real intervention was to exploit the actual constraint. Minor variants were delegated to a trained resource, and technical office was protected from commercial interruptions with two scheduled reception windows. Throughput rose to 8.5 jobs per week and lead time fell to 17 days.

The new cutting centre eventually became useful, because once the technical constraint was elevated, the constraint moved to cutting. This time the company knew it before investing further.

Diagnostic questions

Where does work accumulate in the main flow, physically or digitally, in a stable way?

Is the accumulation point the same as the point people believe is the problem?

Did the last capacity investment reduce total lead time, or only local effort?

If it reduced only local effort, the company probably improved a non-constraint.

Practical implications

A minimal practice is a monthly queue review. Use data the company already has: open files by status, unanswered approvals by function, orders by step, tickets by owner. Ask whether the constraint is still where it was.

Before any capacity investment, use a filter question: does this intervention act on the constraint? If not, it may still be useful, but it should not be justified as a throughput intervention.

Once the bottleneck is identified, respect the sequence: exploit, subordinate, elevate. Money is the last lever. Often the first improvement is simply limiting work-in-progress upstream of the constraint so that the system stops feeding it more than it can absorb.

MARTRO reading

In MARTRO’s reading, the bottleneck is the meeting point between process analysis and decision architecture. In operationally mature SMEs, the constraint often moves upward: from machines and departments to approvals, priorities and founder decisions.

This is why a MARTRO diagnosis does not stop at the physical flow. It also searches for bottlenecks in decision latency, where the queue is made of choices waiting for authorisation rather than material waiting for work. A founder who discovers that they are the bottleneck has usually found the highest-leverage intervention in the company.

When to go deeper

Go deeper before any capacity investment and whenever total lead time worsens while everyone looks very busy. That combination is the classic signature of a misidentified constraint.

Natural next steps are WIP - Work in Progress, operational capacity and the practical guide on how to identify a bottleneck.

Frequently asked questions

Is a bottleneck always bad? No. Every system has a constraint. The problem is not knowing where it is. Well-managed systems sometimes choose where the constraint should sit and subordinate the rest of the flow to it.

How can we find it without consultants or software? Follow the queues for two or three weeks. Count work waiting before each step. The step with a persistently growing queue is the candidate. Then test whether a small increase in its capacity changes total throughput.

Can a person be the bottleneck? In SMEs this is very common. A unique competence, a signature authority or the founder’s attention can be the constraint. In that case the solution usually involves delegation, decision thresholds and role design, not machinery.

Why does it matter to an investor? Because the bottleneck determines real scalability. Two companies with the same EBITDA may have very different value if one knows what will break under growth and the other does not.