Production bottleneck analysis: bu ne anlama geliyor
A bottleneck is the resource that limits total system output. The founding observation of the theory of constraints is that system capacity equals the bottleneck’s capacity, not the average of its stations.
The most common error
“The machine with the highest utilisation is the bottleneck” is a frequent assumption and frequently wrong. A station may show high utilisation because the work arriving at it is irregular — without being the constraint at all.
Four ways to find it
- Queue length. A station with work permanently piling up in front of it is a strong candidate. Measure the average queue over time, not at a single moment.
- Waiting time. If other stations are waiting on this one — for material or for their turn — that is the bottleneck.
- Utilisation rate. Stations running continuously above 85-90%. But do not use this alone; that is the trap above.
- Experimental confirmation. The most conclusive method: temporarily raise the candidate’s capacity (add a shift, add a temporary resource) and measure whether total output rises. If it does not, the bottleneck is elsewhere.
Bottlenecks move
When the product mix changes, the bottleneck relocates. In a month weighted towards product A the constraint may be the paint shop; towards product B, assembly. Bottleneck analysis is therefore continuous measurement, not a one-off project.
After you find it
The theory of constraints prescribes a clear order: exploit the bottleneck (never let it idle, never stop it for breaks, never feed it defective material), subordinate the system to it (other stations run to its rhythm), then elevate it (invest).
Most companies jump straight to the third step. The first two are usually free and typically yield 10-20%.
What the data shows and what the floor shows
Bottleneck analysis can be done at a desk with MES data, but it is incomplete without floor verification. Typical cases the data cannot see:
Operator dependency. A station may run at half speed without a particular operator. Machine capacity data does not show this.
Changeover. Setup time is recorded in most MES systems as non-production time and deducted from capacity, but not treated as a separate constraint in bottleneck analysis.
Quality-driven rework. If 8% of what leaves a station returns to the previous one, that station’s real capacity is lower than it appears.
The buffer in front of the constraint
The theory of constraints says the bottleneck must never be starved. That calls for a deliberate work buffer in front of it — which appears to contradict the general “reduce inventory” principle but does not.
The difference: a buffer in front of the constraint protects produced value; stock anywhere else ties up capital. The right question is not “is there stock” but “is the stock in the right place”.
Investment decisions
Capacity investment in a non-bottleneck station does not raise total output. It only extends that station’s idle time and accumulates inventory.
The bottleneck moves
Solving a bottleneck does not remove it — it relocates it. The station whose capacity was raised is no longer binding; the next one is.
So “we found and fixed the bottleneck” is usually incomplete. The right question is where it moves next and whether that station is ready. Investment decisions change substantially on the answer.
Queue length is the honest indicator
Utilisation misleads: a station can show 95% utilisation while constantly waiting, because the ratio measures running time rather than waiting.
The amount of work piling up in front of it says so directly. Noting which station has work stacked at the end of each shift, for one week, is faster and clearer than most analytical tools.