A roof cleaning robot pays back on skylight work when it covers at least 900 m² per shift on flat to 15° glass, and loses money below roughly 400 m². Between those two numbers everything depends on pad pressure, access logistics and how fast your team can reposition the machine.
Five years of watching these machines move from one-off demos to actual site kit has changed what I would buy. Not the model. The category of decision.
What does a roof cleaning robot actually do on a skylight roof?
It runs a brush or microfibre pad across the glass under suction, driven by two belt track sets or four wheels. Water and dirt get pushed to a recovery edge, or wiped off dry on coated glass.
On a skylight roof the job is different from a vertical curtain wall. The glass sits at 0-30° most of the time, which helps, because gravity is not fighting the suction cups. It also hurts, because leaves, grit and wind-blown sand accumulate in the same plane the tracks need to grip.
In 2025 we tracked a 1,400 m² atrium roof in Guangdong. Two machines, one operator, four hours per clean, twice a month. Manual rope work on the same roof took six hours with three people and a permit. The saving was not the hours. It was the permit queue.
How many machines does a mid-size roof actually need in 2026?
| Roof area | Glass type | Machines | Shifts per clean | Notes |
|---|---|---|---|---|
| under 300 m² | single-pane skylight | 1 | 0.5 | Manual may win on cost |
| 300-900 m² | laminated / tempered | 1-2 | 1 | Watch pad load limits |
| 900-2,500 m² | tempered + coating | 2-3 | 1-1.5 | Spare pad set essential |
| 2,500 m²+ | mixed deck | 3-5 | 2 | Fleet scheduling becomes the job |
The table hides the real constraint. A roof robot needs a water supply and a place to park between cleans. If the only roof access is a 700 mm hatch, a 12 kg machine with a 400 mm body is manageable and a 26 kg unit with a 600 mm body is not. Measure the hatch before you read any datasheet.
Where do skylight fleets fail?
- Pad glazing on laminated glass and leaving the machine still. Two hours of static load at 30 °C softens the PVB interlayer and leaves a visible mark. Move it or lift it.
- Running dry on coated low-e glass. The coating is thin and scratches. Dry wipe only with a fresh microfibre, never the brush.
- Ignoring drainage. A 6 mm standing water film at the valley of a pitched skylight turns into tide marks within an hour of sun.
- Skipping the tether inspection. The tether is rated, but UV degrades webbing. 12 months is the usual replacement interval in the Middle East, 24 months in northern Europe.
None of those are machine faults. All four are deployment faults, which is why a cheap unit with disciplined procedures beats a premium unit left running unattended.
Which roof robot suits which buyer?
Facilities teams with 1,000 m² of skylight and a full-time maintenance crew: buy two units and train one person properly. The training is worth more than the second machine.
Cleaning contractors bidding skylight jobs: buy one unit with a documented duty cycle and rent a second for peak season. Fleet capex on a seasonal contract kills cash flow.
Homeowners with one 4 m² Velux-style skylight: do not buy a roof robot. The economics never land. A pole with a pivot head and a step ladder works, and the risk is lower.
If you are still deciding whether a robot fits your glass at all, read our broader skylight robot reviews before you take a quote. The comparison material there covers pad types, tether ratings and the repair history on units that have done more than 1,000 hours.
What does fleet operation cost per year?
Take a two-machine roof fleet. Pads, 8 sets a year at USD 90-160. Squeegee blades, 4 sets. One motor service per machine at around USD 140. Tether replacement every two years, USD 200 per machine. Labour for daily checks, 25 minutes per machine per clean.
Add it up for a roof cleaned twice a month: roughly USD 1,900-2,600 per machine per year in consumables and service. That is the number to put against a manual crew invoice, not the sticker price.
Our Lingkong K3 review covers what higher-load machines handle on sloped glass, if your roof sits above 20°.
Key Takeaways
- A roof cleaning robot is viable above roughly 900 m² per shift; below 400 m² the maths fails.
- Hatch size and parking space constrain the machine more than suction specs do.
- Static pad load on laminated glass at high ambient temperature is the most common damage claim.
- Budget USD 1,900-2,600 per machine per year for consumables and service in 2026.