Find the bottleneck that actually limits customer output.

Define the output you need

Choose a product family, route and observation period. Record demand and good completions at the output boundary. Keep machine cycles, rework and shipped units separate so the analysis compares like with like.

Combine capacity with event evidence

Compare effective capacity across steps, including relevant setup, downtime and yield. Then inspect what the processes actually waited for.

A persistent upstream queue and downstream starvation can support a bottleneck hypothesis. Buffers, product routes and release rules can change that pattern, so confirm it by following events rather than reading one snapshot.

A capacity calculation gives you a candidate to check

Illustrative example: one product follows cutting, assembly and packing. Each step has 420 net scheduled minutes per day. The effective minutes below already exclude expected local losses; do not multiply another OEE factor into them. Demand is 100 accepted units per day, with no yield loss in this simplified case.

Persistent and shifting constraints need different observations

If assembly repeatedly limits output across comparable days, investigate its recurring capacity losses first. Packing may be starved while waiting for assembly. Assembly may also be blocked when a full downstream buffer prevents a handoff. Separate these states in the event record before choosing an action.

With a different product mix, breakdown pattern or staffing plan, the limiting step can move. Persistent and shifting are useful descriptions, not an exhaustive two-type classification. Compare more than one observation window and recheck the whole route after a change.

Distinguish a symptom from a limiting loss

The station where a stop appears may be responding to a loss elsewhere. Trace the sequence and check whether a buffer protected output during the interruption.

Test the candidate across ordinary disruptions

Repeat the comparison for important products, shifts and downtime patterns. If a simulation is used, first check that its baseline represents the observed flow closely enough for the decision.

1. Follow a job

   Identify where it waits and which event releases it.

2. Form a hypothesis

   Name the resource or rule thought to limit output.

3. Test a change

   Compare the flow with and without a specific loss or restriction.

4. Check side effects

   Review quality, inventory and downstream limits.

5. Choose the next check

   Use a safe trial or better observations to resolve remaining uncertainty.

Recheck when conditions change

Demand, mix and reliability can move the bottleneck. Keep the evidence period with the finding and revisit it after a substantial change. A successful local fix is useful only if its effect on the wider flow is understood.

Manage the limiting step without chasing a numbered list

In simple terms, a bottleneck is the resource or operating rule that limits the whole flow’s good output under the conditions being studied. A busy station or a long queue is a clue, not sufficient proof.

A useful four-point operating summary is to release work at the constraint’s feasible pace, protect its useful capacity, avoid expecting extra non-constraint capacity alone to raise output, and test whether increasing the constraint’s capacity raises system output. This is a working summary, not a separate universal standard. Safety, maintenance and quality requirements still apply.

Theory of Constraints uses five focusing steps: identify the constraint, make better use of it, align the rest of the system to it, increase its capability if needed, then repeat when it moves. TOCICO teaches that named five-step model. Do not drop the last step just to fit a four-item search question.