Laser Cutting Workflow: Measuring and Removing the Real Bottleneck

Cutting speed is an obvious place to look, but it is only one of roughly nine stages between a releasable order and the next operation. Until those stages are timed separately, there is no reason to assume cutting is the constraint. Fixing the wrong stage — buying a faster machine when the real delay is a late drawing or a full unloading table — spends money without changing the number of parts that ship. The fix starts with measuring, not guessing.

The stages to time, and where each one starts and stops

Pick a part family that runs often enough to be representative and time every stage across several real jobs, not one. The definitions matter more than the stopwatch: two people timing "setup" will get different numbers unless they agree where setup begins and ends. A workable set of definitions:

Stage	Starts when	Ends when	What it captures
Queue	The job is releasable — drawing, material and quantity confirmed	Someone starts work on it	Waiting that has nothing to do with machine capacity
Program	The programmer opens the job	The nest is checked and saved ready to load	CAD/CAM, nesting, and any chasing of missing information
Setup	The previous job is off the machine	The machine is ready to start this job	Consumable changes, parameter and gas selection, fixturing
Load	Material is brought to the machine	Material is clamped and the datum confirmed	Crane, forklift or shuttle-table time, including waiting for the handling equipment
Cut	Cycle start	Cycle end	Beam-on time plus in-cycle moves — the number everyone already tracks
Unload and sort	Cycle end	Parts are separated from the skeleton and in a bin or on a pallet	Part removal, skeleton handling, sorting, tabbing
Inspect	Parts are presented for checking	Parts are released or held	First-piece and in-process checks, including waiting for the person who does them
Rework / re-cut	A part is held	A replacement or corrected part is released	Everything a failed part costs, including the re-queue
Move to next operation	Parts are released	The next operation starts on them	Transport and the wait at bending, welding or shipping

Log start and end times for each stage on a handful of real jobs rather than estimating from memory. Memory-based estimates can understate waiting, handling and inspection because those stages are less visible than machine cycle time.

A one-shift, one-job worksheet

Copy this into a spreadsheet, one row per job, and fill it from timestamps written on the traveller or a clipboard at the machine. Elapsed time is what matters here (clock time from start to end of the stage), not labour hours — a part that sits in a queue for six hours consumed six hours of lead time even though nobody touched it.

Job / part	Queue	Program	Setup	Load	Cut	Unload & sort	Inspect	Rework	Move to next op	Total elapsed	Notes (what it was waiting on)

The Notes column matters as much as the numbers. "Waiting for forklift", "drawing missing material grade", "inspector on another job", "bending queue full" — written at the time, not reconstructed later — is what turns a column of numbers into a diagnosis.

A hypothetical worked example

Say a fabrication manager times five identical bracket jobs over three weeks and logs the following elapsed hours per stage. These are hypothetical figures for illustration only, not a benchmark:

Job	Queue	Program	Setup	Load	Cut	Unload & sort	Inspect	Move to next op
1	6.0	0.3	0.4	0.2	0.6	0.3	0.2	2.5
2	1.5	0.3	0.4	0.1	0.6	0.3	0.2	3.0
3	8.0	1.2	0.5	0.2	0.6	0.3	0.2	2.0
4	2.0	0.3	0.4	0.5	0.6	0.3	0.6	2.5
5	5.5	0.3	0.4	0.2	0.6	0.4	0.2	4.0

Cut time is the smallest and most stable number in every row. Queue and "move to next operation" are both larger and far more variable, and Job 3's inflated program time lines up with a note that the drawing arrived without a material grade and had to be confirmed before nesting could start. Job 4's load time doubled because the forklift was on the receiving dock. None of that shows up if the shop only logs cut time, which is what a machine-utilisation report captures by default. The manager in this example would be justified in looking at queueing and the handoff to bending before considering a faster machine, because cutting is already a small share of the elapsed time per job.

This is the pattern worth checking for on your own data: does the stage with the biggest total, or the widest swing between jobs, match the stage you would have blamed from memory? If not, the memory-based answer was wrong and the fix belongs somewhere else.

Questions to ask of the data

Once the worksheet has a few weeks of jobs in it, work through a short list of questions rather than relying on the averages:

• Where does work-in-process accumulate? Walk the cell and count: sheets waiting to be loaded, skeletons waiting to be cleared, cut parts waiting for inspection, bins waiting for bending. WIP accumulation is a useful physical signal to compare with the timing data.
• Which step prevents the next job from starting? When the machine is idle, what is it waiting for — material, a program, an operator who is unloading the last job, an inspector?
• Is the machine waiting, or is something waiting for the machine? These are opposite problems. A machine that waits is underfed; a queue that waits is undersized capacity or poor sequencing.
• Does a faster cut actually increase completed parts through the whole cell? If unloading, sorting or bending cannot absorb more output, a faster laser produces a larger pile, not more shipped parts.
• Is the variation in a stage bigger than its average? A stage that takes 20 minutes every time is a characteristic to plan around. One that takes 20 minutes on one job and three hours on another is telling you something upstream is unstable, and that variability — not the average — is often the more useful thing to fix first.
• How often is a job released before it is actually ready? Every "waiting on drawing / material / confirmation" note is a release that happened too early.

How much data is enough

A small sample can reveal a possible pattern, but use a broader sample before making a capital, staffing or scheduling decision. Before committing capital or reorganising a shift around a conclusion, extend the log across a broader sample of the same part family, different days of the week and different operators, and confirm the pattern holds. A constraint that only appears on Mondays after a weekend shutdown is a different problem — and a different fix — from one that appears on every job regardless of when it runs.

Capacity constraint, scheduling constraint, or information constraint

Not every slow stage is a capacity problem. Three different things look similar on paper:

• A capacity constraint — the stage is consistently full; the machine, the operator or the equipment doing that step is at its physical limit and there is no slack to absorb more work.
• A scheduling constraint — the resource has spare time, but jobs arrive in the wrong order or several compete for it at once, creating queues a true capacity shortfall would not.
• An information constraint — the work cannot start, not because the resource is busy, but because something it needs (a drawing, a material confirmation, a nest, an approval) has not arrived.

The distinction matters because the fix is different in each case. Adding capacity to a scheduling or information constraint does not remove the queue; it gives the queue somewhere faster to sit and wait.

Telling a constraint from a symptom

Suspected cause	Quick test	What it tells you
Capacity constraint	Track the stage's utilised hours against its available hours over a full week.	If utilisation sits near the available hours with little slack, added capacity is the correct lever.
Scheduling constraint	Check whether the resource has idle blocks even while a queue exists in front of it.	Idle time next to a queue points to sequencing or batching, not a shortage of capacity.
Information constraint	Trace what the stage was waiting on the moment it restarted — a document, a confirmation, an approval.	If the same missing input recurs across jobs, the fix is upstream of the floor entirely.
Handling constraint	Compare load and unload time with cut time, and note how often the crane or forklift was elsewhere.	If handling routinely exceeds cut time, the laser is waiting on material movement, not on its own speed.

A machine that looks starved for work because programs are not ready is not undersized; it is underfed. Buying a second machine in that situation adds a second starved machine.

Office-side causes

Some of what looks like a shop-floor bottleneck traces back to the office:

• Late or incomplete drawings — work released to the floor before it is ready to run, creating a queue at whichever stage discovers the gap.
• Revision churn — a part re-released after cutting has started, forcing rework or reprogramming.
• Incomplete nesting data — missing grade, thickness or quantity that stalls programming until someone chases it.
• Missing material — a job scheduled against stock that has not arrived or been confirmed, discovered only when someone goes to pull it.

None of these show up if you only time the machine. They show up when queue and program are timed alongside cut and unload, which is why those stages belong in the same measurement exercise rather than being treated as a separate "office problem".

A short improvement sequence

Once the data points to a real constraint, work through it roughly in this order:

1. Fix information gaps first — drawing completeness, revision control and nesting data — because they may be solvable without adding equipment and they inflate every downstream stage's measured time until they are gone. The companion article on building a reliable CAD-to-cut workflow covers this in detail.
2. Fix sequencing next — grouping jobs by material and thickness, and releasing work only when it is genuinely ready — before assuming more equipment is needed.
3. Fix handling if load and unload are consistently competing with cut time; the material handling article covers when that justifies automation and when a second pallet or a dedicated forklift slot is enough.
4. Address the confirmed capacity constraint last, whether that is cutting, handling or a downstream operation. If bending consistently backs up behind cutting, the constraint may not be the laser at all; see the press brake range and the press brake buying guide for what a matched bending step looks like.

Measuring first means the investment, when it happens, goes to the stage the data actually points to.