×
Oct 28,2026
I signed three purchase orders for laser cleaners in two years. Two of them were mistakes. Here is the invoice-level detail, so you can skip the tuition.
This is the buyer's side of the table: five traps, what each one actually cost, and the one question that defuses it. No theory. Every figure below comes from my own purchase orders, my own time sheets, or my own electricity and freight bills.
If you only read one table, read this one. It is the whole article compressed into the arithmetic that matters — what the paper promised against what the shop delivered.
| Trap | What the sheet said | What the shop gave me | What it cost |
|---|---|---|---|
| 1. Cleaning width | "20 mm cleaning width" | 8 mm at a hand-held speed, on real rust | $18,400 lost margin on one bid |
| 2. Duty cycle | "100% duty, continuous" | 74% output at 38 °C, 4 shutdowns a shift | $5,300 in idle labour |
| 3. Consumables | Not mentioned | Windows and a focus lens, every quarter | $2,300 in year one |
| 4. Warranty | "2-year warranty" | Source only; the parts that fail are 6-month items | $1,900 freight + 5 weeks |
| 5. The demo | Video of a clean 1.5 mm plate | My work is 8 mm plate under mill scale | $3,420 subcontracted |
Table 1 — Five traps, five invoices. The machine cost $9,800. The traps cost $31,320.
That ratio is the reason I am writing this. Nobody loses money on the sticker price. You lose it in the gap between a measurement taken on a laboratory coupon and a measurement taken on your floor, at 4 p.m., on a Friday, with a operator who has been holding a 1.6 kg head for six hours.
My first machine was sold as a 500 W unit with 20 mm cleaning width. Both figures are defensible. The 20 mm is real — on a polished plate, at 0.3 m/s, with a single pass over a film of fingerprint oil. My job was 1,200 m² of structural steel with six-month rust and two coats of old alkyd paint.
I bid the job at the brochure rate. I delivered at the shop rate. −$18,400
The arithmetic is not complicated, which is what made it so annoying to discover late. Throughput is line width × travel speed × overlap factor. Everything else — watts, spot size, cooling — only exists to make that product possible.
| Coating on the plate | Brochure width | Production width (hand-held, 50% overlap) |
Throughput I actually got |
|---|---|---|---|
| Oil film, <5 µm | 20 mm | 18–20 mm | 4.9–5.4 m²/h |
| Flash rust, days old | 20 mm | 14–16 mm | 3.8–4.3 m²/h |
| Medium rust, 6 months | 20 mm | 8–9 mm | 2.2–2.4 m²/h |
| Old paint, 2 coats | 20 mm | 6–7 mm | 1.6–1.9 m²/h |
| Mill scale, hot-rolled | 20 mm | 5–6 mm | 1.3–1.6 m²/h |
Table 2 — Same machine, same 500 W, five coatings. A 3.4× spread in output from one unchanged datasheet line.
Note what happened to the width. It did not shrink because the machine lied. It shrank because holding 20 mm at 0.3 m/s delivers roughly 4 J/cm², and six-month rust needs 35–45 J/cm². To reach that fluence the operator does the only thing available to him: he slows down and narrows the line. The machine's limit was never width. It was energy per square centimetre, and width is the variable that gets sacrificed to meet it.
My second machine is air-cooled, and the datasheet says 100% duty cycle, continuous operation. In February that was exactly true. In August, with the shop at 38–41 °C and the roller door open, the same machine folded back its output and shut down four to seven times per shift.
Forty-one hours of stopped work across one month, at a shop rate of $130/h. −$5,300
| Shop ambient | Available output | Thermal shutdowns per 8 h | Effect on my day |
|---|---|---|---|
| 22–25 °C | 100% | 0 | Rated throughput, no interruptions |
| 30 °C | 96% | 0 | Not measurable |
| 35 °C | 88% | 1 | One coffee break you did not schedule |
| 38 °C | 74% | 4 | Every long pass interrupted |
| 41 °C | 62% | 7 | Shift plan abandoned by 2 p.m. |
Table 3 — Air-cooled 500 W unit, measured over six weeks with a logging thermometer beside the intake. Water-cooled units hold their rating to 45 °C and cost about $600 more.
The lesson is not "buy water-cooled". It is that the cooling system is a specification, not an accessory, and it is the only spec that changes with the weather. A shop in Malaysia and a shop in Norway can buy the same laser source and get different machines. If your summer peaks above 35 °C and you are buying air-cooled, budget for the derating in your throughput calculation — or put the chiller on the order.
Nothing in my quotation mentioned a protective window. It is a $48–120 disc of optical glass that sits between the lens and the world, and it exists so that the spatter eats it instead of a $700 focus lens. On heavy rust it lasts 120–300 hours. On paint and mill scale, less.
Year one, 1,100 operating hours: seven windows, one focus lens, two nozzles, plus four days of waiting for parts. −$2,300
| Consumable | Unit price | Life (heavy rust) | Cost per 1,000 h |
|---|---|---|---|
| Protective window | $48–120 | 120–300 h | $160–1,000 |
| Focus lens | $380–900 | 800–1,500 h | $250–1,100 |
| Nozzle / air knife | $25–60 | 400–800 h | $30–150 |
| Deionised water (water-cooled) | $2/L | 2 L per 400 h | $10 |
| Total, realistic | — | — | $450–2,350 |
Table 4 — The line item missing from most quotations. On 1,100 h/year, this is a 5–24% annual addition to the machine's cost.
There is a second, quieter cost inside this one: the four days you wait for a window. If the machine is on the critical path, that is four days of a contract deadline, and no supplier refunds it. I now keep two spare windows on the shelf. They cost $180 and they have saved me two weekends.
"2-year warranty" sounds like two years of protection. Mine covered the laser source for 24 months, the galvo head for 6, the handheld cable for 3, and the optics for none — they are classified as consumables.
The source is the most reliable component in the machine. The galvo, the cable and the optics are the three things that actually break. −$1,900 + 5 weeks
| Component | My warranty | What actually fails | Freight, each way |
|---|---|---|---|
| Laser source | 24 months | Rarely — 20,000+ h MTBF | — |
| Galvo scan head | 6 months | Bearings, mirror coating | $220–400 |
| Handheld cable | 3 months | Flex fatigue at the strain relief | $150–300 |
| Optics / windows | None | Spatter, thermal crazing | $40–90 |
| Chiller | 12 months | Compressor start capacitor | $180–350 |
Table 5 — Read the warranty as five separate warranties. The headline number is the one for the part you will never claim on.
The demo video showed a 1.5 mm mild steel plate, brushed clean, with a bloom of orange flash rust. It came off in one pass at what looked like walking speed. I ordered on the strength of it.
My work is 8 mm hot-rolled plate with mill scale and two coats of alkyd. The same machine, on my plate, ran 34% below the throughput I had planned for, and I subcontracted 380 m² at $9/m² to hold the deadline. −$3,420
| Variable | The demo plate | My production plate | Throughput effect |
|---|---|---|---|
| Thickness | 1.5 mm | 8 mm | Heat sink, −12% |
| Surface | Brushed, uniform | Mill scale + paint | −28% |
| Geometry | Flat coupon | Stiffeners, welds, corners | −9% |
| Operator | Sales engineer, fresh | My welder, hour six | −7% |
| Combined | 100% | — | 66% of demo rate |
Table 6 — Four differences, none of them dishonest, that turn a demo into an overrun. The 34% gap is what I subcontracted.
Five traps taught me one habit: I no longer evaluate a quotation. I evaluate a completed specification, and I refuse to compare prices until every row below has an answer. It takes a supplier twenty minutes and it has saved me more than the machine cost.
| Ask exactly this | Acceptable answer | Walk away if… |
|---|---|---|
| Cleaning width at what speed, on what coating? | Three numbers: width, m/s, coating + thickness | A single width figure |
| Duty cycle at 40 °C ambient? | Percentage + derating curve | "100%, no problem" |
| Consumable cost per 1,000 h? | Itemised, with part numbers and lead time | "Consumables are cheap" |
| Warranty by component? | A table, plus who pays freight | One headline number |
| Will you test my coupon on video? | Timer visible, pass count stated | "Our standard demo is the same" |
| Throughput in m²/h on my sample? | A measured figure with overlap stated | A range with no test behind it |
| Spare parts held in stock? | Windows, lens, cable, nozzle — on the shelf | "We order from the factory" |
Table 7 — Seven rows. If two or more come back vague, the price is not the number you should be comparing.
The machine was never the problem. I bought a good machine twice and lost money both times because I compared numbers instead of conditions. Every trap above disappeared the moment I started asking for the same measurement, on my metal, at my ambient temperature, at my operator's speed — which is a two-line email and a courier envelope.
If you remember one line: a datasheet describes a machine in a laboratory; your shop is the experiment. The only number worth trusting is the one measured on your own offcut, and it costs you a stamp to get it.
We now send every buyer a test report written in exactly the format of Table 7 — coating, thickness, ambient, speed, pass count, m²/h, consumables per 1,000 h — and we send it before the quotation, not after. Some buyers find that the honest number is smaller than the brochure. They buy anyway, because they can plan around a small number and they cannot plan around a surprise.
The 500 W unit from Trap 1 — a good machine sold with a bad number. Two spare windows on the shelf cost $180 and ended the downtime.
EN
