Bottleneck management: finding the resource that sets output and running it properly

Locating the constraint with evidence rather than reputation

Every plant has a folk answer to this question and it is often wrong, usually naming the noisiest or most complained-about machine. Better evidence is physical and cheap to gather: where does work queue and stay queued, which resource is asked for overtime most often, which one do supervisors expedite around, and where does the plant hold protective stock. Add the arithmetic check of loading against available hours for the current order book. Where the physical evidence and the arithmetic disagree, believe the queue, because it reflects the plant that exists rather than the routings that were entered.

Hours lost at the constraint are lost to the business

This is the operating principle that changes behaviour. Setting up the constraint during a break, taking it down for maintenance during a production window, running trial or sample work on it, letting it wait for an operator on another job, or feeding it material that will later be scrapped are all decisions that cost plant output. Practical countermeasures include inspecting components before they reach the constraint, staffing it through breaks, giving its changeovers priority for the best setters, and scheduling its maintenance around production. None of these are complicated; they simply require someone with authority to insist.

Buffering the constraint without flooding the plant

The constraint should never starve, which means holding a deliberate quantity of work in front of it, sized by how long upstream disruptions typically last. That buffer is a designed feature and it needs a target level and someone watching it. The failure is generalising the idea and buffering everywhere, which fills the plant with work in progress, hides problems and lengthens delivery. Keep protective stock in the few places that protect output — before the constraint, and often before final assembly or dispatch — and keep the rest of the flow lean enough that a problem is visible when it happens.

When the bottleneck moves with the product mix

In mixed plants the constraint is not fixed. A shift towards items with more machining moves it to the machine shop; a shift towards items needing cure or test moves it downstream. Managing this requires monitoring loading across the small set of candidate resources rather than assuming last quarter's answer still holds, and it requires the commercial team to understand that different work loads the plant differently. A wandering constraint also undermines any improvement programme aimed at a fixed target, since capacity added to a resource that is no longer limiting produces nothing at all.

Deciding whether to elevate the constraint or live with it

Once the constraint is being run well, the choices are to add capacity, offload work to another route or a subcontractor, or accept the limit and manage demand. The evidence for spending is recorded loading against available hours over a period long enough to show the pattern, plus proof that the easy recovery has already been done, because buying a machine to relieve a resource that loses many hours to changeovers and stoppages simply purchases two poorly run machines. The decision also needs a view of where the constraint will move next, since relieving one resource always promotes another.

Frequently asked questions

What is the difference between a bottleneck and a capacity-constrained resource?

A bottleneck cannot meet the demand placed on it at all, so work accumulates permanently in front of it. A capacity-constrained resource has just enough capacity in total but loses its position whenever it is disrupted or badly sequenced, so it causes late delivery intermittently rather than continuously. The management response differs: the first needs more capacity or less demand, the second needs better scheduling, protected setups and disciplined maintenance.

Should we run the non-constraint areas at full utilisation?

No. Non-constraint resources should produce what the constraint and the schedule need, when they need it, and then stop or change over. Running them flat out builds stock the plant cannot convert, consumes material early, occupies space and hides quality problems until much later. Expect this to be uncomfortable for supervisors judged on utilisation, which is why the measures for those areas have to change at the same time as the working practice.

How do we protect the constraint from bad material reaching it?

Move the relevant quality checks upstream so defective components are found before they consume constraint hours. That may mean inspecting at the previous operation, at goods receipt for bought parts, or with a quick in-process check by the operator feeding the constraint. The economics are straightforward: the cost of the earlier check is small compared with the output lost when the constraint processes a part that is then scrapped, and the loss is repeated every time.