All articles

How to Find and Fix a Bottleneck in Production

By Davina · Published

Key takeaways

  • A bottleneck in production is the station whose capacity falls below demand, capping the throughput of the entire line regardless of how fast other stations run.
  • Four signals identify a constraint: WIP piling up upstream, starvation downstream, utilisation near 100%, and cycle time exceeding takt time.
  • A five-step analysis (map, measure, compare to takt, verify with a system test, project the next constraint) runs on a spreadsheet and one site visit.
  • Fixing a bottleneck moves the constraint to the next-worst station, so calculate remaining headroom against takt time before quoting further work.
  • Offloading tasks to non-constraint stations and cutting changeovers with SMED raise throughput without new capital equipment.

A bottleneck in production is the station with the lowest capacity relative to demand, and it caps the throughput of the entire line no matter how fast the other stations run. This guide walks through definition, detection, analysis, and resolution using one production line and one set of numbers carried end to end, including where the constraint moves after the fix.

What is a bottleneck in a production process?

A bottleneck in production is the step whose capacity is lower than the demand placed on it, which caps the throughput of the whole line. No matter how fast the other stations run, output cannot exceed what the constraint can process. This follows the Theory of Constraints, which holds that every system has at least one constraint governing total throughput, as summarised in this operations-management reference on the Theory of Constraints.

Two signs mark a constraint: work piles up in front of it (WIP accumulation), and downstream stations starve behind it. You can see both patterns worked through with real numbers in the workflow bottleneck calculator example.

What does bottleneck mean in production versus other constraints?

A bottleneck in production is a resource whose capacity is lower than the demand placed on it, so it caps throughput for the whole line. Not every constraint is a bottleneck: a policy (a batch-size rule), a market (not enough orders), or a material shortage can also limit output, but only a capacity constraint inside the process is a bottleneck. Inefficiency is different again. A slow changeover on a non-constraint station wastes time without cutting throughput, so fixing it changes nothing system-wide. Scope work by asking which constraint actually sets output; scoping that question first is cheaper than running simulation before you know the payoff (https://improvefirst.com/learn/bottleneck-payoff-before-simulation).

How do you spot a bottleneck in a production line?

Detection comes down to four signals: WIP piling up in front of one station, starvation downstream of it, utilisation pinned near 100%, and a cycle time that exceeds takt time. If you see two of these at the same station, you have found your constraint.

Reading WIP and queue signals

Walk the line and look for the buffer that never drains. WIP accumulates immediately upstream of the constraint because the station consumes slower than the stations before it feed it. Downstream, the opposite appears: operators and machines waiting on parts, short idle gaps between cycles. The station itself will show near-total utilisation, since any downtime there directly costs throughput. A quick way to scope this before quoting deeper work is to map queue locations and sizes, then test whether the queue sits at the same point every shift (bottleneck software guide).

Comparing cycle time to takt time

Takt time is available production time divided by customer demand (Lean Enterprise Institute glossary). Any station whose cycle time runs longer than takt sets your system capacity. Measure each station over several cycles, not one, and rank them against takt: the worst offender is your bottleneck in production.

How do you run a production bottleneck analysis step by step?

A production bottleneck analysis has five steps: map the process, measure cycle times, compare capacity to demand, verify with a system test, then project where the constraint moves. Run them in order and you can scope a client engagement in a single site visit.

Step 1: map the process and measure cycle times

Draw the line from raw material to finished goods, then time each station over at least 20 cycles. Record cycle time per station, not averages only: the spread matters. The station with the longest median cycle time is your suspect.

Step 2: compare capacity to demand

Convert customer demand into takt time (available production time divided by demand). Any station whose cycle time exceeds takt time cannot keep pace and becomes the constraint. Also check WIP: a growing backlog ahead of one station confirms it.

Step 3: test whether improving the suspect changes system output

The critical step. Speed up the suspect station and measure total line throughput, not the station's own output. If system throughput rises, you found the constraint. If it doesn't, the constraint sits elsewhere, often upstream in material supply or downstream in changeovers. A tool like the Constraint Impact Calculator (https://improvefirst.com) runs this test before you commit to a fix.

Step 4: predict where the constraint moves next

Fix one constraint and the next-worst station becomes the new limit. Calculate each remaining station's headroom against takt time and tell the client which station will cap throughput next. That projection is what turns a one-off fix into a quoted improvement programme.

What is an example of a bottleneck in production?

The line and the demand

Take one line with five stations. A client needs 480 units per eight-hour shift, so takt time is 60 seconds. Measured cycle times are: Station 1 at 50 s, Station 2 at 65 s, Station 3 at 55 s, Station 4 at 40 s, Station 5 at 45 s. Station 2 exceeds takt; everything else has slack.

Detection: where WIP piles up

Run the line for a shift and watch the queue in front of Station 2. It starts near zero and grows steadily, because Station 2 can only finish 3,600 ÷ 65 ≈ 55.4 units per hour against a demand of 60. That is the demand gap: about 4.6 units per hour, or roughly 37 units short by shift end. Little's Law states that average WIP equals throughput multiplied by lead time, so as throughput through Station 2 falls below arrival rate, the queue in front of it must grow without limit.

The pile-up is not a scheduling mistake; it is arithmetic. Stations 4 and 5 starve, and the shift ends with a backlog that overtime only patches. This is the same logic behind queueing in office processes, as in our invoice processing capacity example.

The fix and the recalculation

Station 5 runs 15 seconds under takt, so move one 15-second task from Station 2 to Station 5. Station 2 drops to 50 s, throughput rises to 3,600 ÷ 55 ≈ 65.5 units per hour, and the line now clears 480 units with margin.

Where the bottleneck moves next

Station 3 at 55 s becomes the new constraint, matching Station 2's new cycle time. Re-measure before quoting further work.

How to reduce bottleneck in production without new capital equipment?

Offload work to non-constraint stations

Every task at the constraint that could run elsewhere is throughput you are burning. Walk the line and list what the bottleneck station does: prep, inspection, kitting, labelling. Move any of it upstream or downstream where the cycle time has slack. In the worked example above, moving final visual checks off the constraint station freed roughly 20 seconds per unit, lifting line throughput with no spend. Also check whether rework loops return to the constraint; catching defects before the bottleneck protects its capacity.

Cut changeover and batch-size waste at the constraint

Changeover time at the constraint is pure lost throughput, so attack it with SMED: externalise setup steps so they happen while the line still runs, and standardise what remains. Then review batch sizes. Large batches look efficient but inflate WIP and lead time, and they can leave a downstream demand gap when the constraint is tied up on the wrong item. Smaller batches at the constraint, paired with faster changeovers, usually raise effective capacity even when raw cycle time per unit rises slightly.

How to avoid bottleneck in production when demand changes?

A constraint that holds at one demand level can disappear at another. Re-run the capacity comparison whenever the forecast shifts, and check where the constraint moves next.

Re-check capacity against each demand level

Take the demand gap from your worked example and test three scenarios: current demand, forecast demand, and a stretch case. For each, recalculate takt time and compare it against the cycle time of every station. The station with the smallest margin changes as demand rises, so quote the client on the work that fixes the next constraint too, not just the current one.

Demand scenarioTakt timeBinding stationMargin
Current6.0 minStation 30.4 min
Forecast5.5 minStation 30.1 min
Stretch5.0 minStation 2-0.2 min

When a spreadsheet stops being enough

A spreadsheet capacity model handles two or three scenarios cheaply, which suits a fixed-fee scoping exercise. For repeat engagements or shifting demand, a purpose-built tool saves the rebuild. ProcessModel documents simulation of process capacity and constraint behaviour when you need to model variability, not just averages. A lightweight calculator sits between the two: fast enough for every re-run, structured enough to hand the client.

Which tool should you use for bottleneck scoping?

For a pre-quote scoping visit, a spreadsheet capacity model is the honest baseline: you already know takt time, cycle times, and WIP arithmetic, and the client can audit every cell. Its weakness is that it shows one snapshot, so testing whether the constraint moves after a fix means rebuilding formulas each time.

Improve First's Constraint Impact Calculator is built for exactly that step: identify the limiting station, model what a fix does to total throughput, and see where the constraint shifts next. DO.Impact and FlowVisual cover similar ground for lighter engagements. ProcessModel suits deeper simulation work, and KaiNexus fits ongoing improvement programmes rather than scoping.

The test worth applying to any tool, including the Calculator: can you reproduce your worked example's numbers, before and after the fix, and defend them to the client? If not, stay with the spreadsheet.

FAQ

Can a bottleneck in production move to another station after you fix it?

Yes. Fix the current constraint and the next-worst station becomes the new limit. In the worked example, moving a 15-second task off Station 2 left Station 3 at 55 seconds as the new constraint, matching Station 2's improved cycle time. Calculate each remaining station's headroom against takt time so you can tell the client which station caps throughput next.

Is the station with the lowest capacity always the bottleneck in a production line?

Not always. The constraint is the station with the lowest capacity relative to demand, and a station can have the longest cycle time while still sitting under takt. Verify with a system test: speed up the suspect and measure total line throughput. If system output does not rise, the real limit sits elsewhere, often in material supply or changeovers.

How much WIP is too much when looking for a bottleneck?

There is no fixed number. The signal is a buffer that never drains: WIP accumulating immediately upstream of one station while downstream stations starve. Little's Law ties average WIP to throughput and lead time, so a queue that grows every shift means arrival rate exceeds that station's capacity. Map queue locations and check whether the pile-up sits at the same point every shift.

Do I need simulation software to do a production bottleneck analysis?

No. The five-step analysis (map, measure, compare to takt, verify with a system test, project the next constraint) runs on a spreadsheet and one site visit. Simulation earns its keep when you need to model variability rather than averages, or when repeat engagements make rebuilding formulas wasteful. For a pre-quote scoping exercise, a spreadsheet the client can audit is the honest baseline.

What is the difference between a bottleneck and a constraint in lean terms?

In lean terms, a constraint is anything limiting system output: a policy such as a batch-size rule, a market with too few orders, or a material shortage. A bottleneck is the narrower case, a capacity constraint inside the process whose capacity sits below demand. Only bottlenecks respond to rebalancing work or cutting changeovers; policy or market constraints need different fixes entirely.