A skylight cleaning robot covers 90-180 m2/h on a real dirty roof in 2026, against a spec-sheet claim of 300-400 m2/h. The gap comes from dirt level, pitch, tank refills and tether moves, not from the robot being slow.
Vendors quote peak speed on a clean, flat pane. Your roof is neither. Knowing the real figure lets you schedule crews properly instead of promising a one-pass clean that takes two days.
What cuts the hourly rate?
| Factor | Effect on m2/h |
|---|---|
| Heavy dirt, first pass | -40% to -55% |
| Pitch over 25 degrees | -15% to -25% |
| Tank refills every 300-500 m2 | -10% to -20% |
| Tether repositioning | -5% to -15% |
| Clean, flat, second pass | near spec |
Run the maths on a 900 m2 mall canopy. At a spec 350 m2/h you would book three hours. Real-world at 140 m2/h, with two tank refills and four tether moves, the job runs about seven hours. That is a full shift, not a morning.
How does battery runtime interact with speed?
Battery and speed trade against each other. A Lingyun Y3 running at high speed drains its pack in roughly 90-120 minutes; dropping to a moderate speed extends runtime past two hours. On a big roof, the slower setting usually finishes the same area in similar total time, because you swap fewer batteries. Two packs, rotated, cover a full day on most commercial roofs.
Does tank size or speed limit the job?
Tank size usually binds before the battery does. A 15-20 L tank holds enough water for roughly 300-500 m2 at 30-45 ml/m2. Above that you refill mid-cycle. If your water point is far from the roof access, refills cost more time than the robot’s speed ever saves.
Who should not chase high speed?
Sites cleaning coated or low-E glass. Faster speed with higher brush pressure strips coatings and voids warranties. Run slower, at the supplier’s brush-pressure figure, and accept a lower m2/h. Also avoid high speed on freshly installed glass during the first months, when construction dust is abrasive.
One concrete trap: quoting hours based on spec-sheet m2/h in a tender. You will lose the argument at the first progress meeting. Quote a range and measure your own first pass. Talk to the team through contact for realistic planning figures.
How do you plan a cleaning schedule from the real rate?
Work backwards from the glass area and the window you have. If you can only clean between 11 pm and 4 am, that is five hours, minus 45 minutes setup and water. A 400 m2 roof at 130 m2/h fits with margin. A 900 m2 roof does not fit in one night, so split it across two nights or move maintenance to weekly light cycles instead of one monthly pass.
For busy roofs, weekly light passes beat a monthly deep clean. Each pass is faster because dirt never binds, water use drops, and the schedule fits tighter windows. The annual throughput is similar, but the peaks disappear.
What slows a robot more than anything?
Repositioning, not cleaning. Every tether move, water refill and section change costs fixed minutes that do not scale with area. On a badly laid-out roof, setup can consume 25-35% of a shift. The fix is layout, not a faster robot: put the water point near the roof access, anchor the tether so it clears fixed structure, and split the roof into two clean zones with their own anchor and draw point.
| Roof layout | Setup share of shift |
|---|---|
| Poor: far water, one anchor | 25-35% |
| Average: reachable water | 15-25% |
| Good: two zones, local draw | 8-15% |
Does robot speed ever matter more than layout?
Rarely. Peak speed only helps on a large, flat, clean roof where the robot can run long uninterrupted passes. On most real roofs, shaving setup time buys more than raising top speed. Buyers fixated on spec-sheet m2/h often pay for speed they can never use because the layout caps throughput first.
Does cleaning frequency change the effective speed?
It changes the effective effort, which is what matters. A monthly pass on moderately dirty glass runs slower than a weekly pass on lightly soiled glass, because the first pass has to loosen and lift more dirt. Moving to weekly light cycles can raise the sustained m2/h enough to offset the extra visits, and it keeps water use lower per square metre over the year.
The point for planners: do not compare a monthly deep clean with a weekly light clean on the same m2/h number. They are different jobs. Compare annual hours, annual water and annual cost for the same clarity outcome.
How do you find your own real rate?
Measure the first three cycles rather than trusting one. Record the glass area, the elapsed working time excluding setup, and the water used. The average of three cycles is your planning figure, and the spread between them tells you how variable the roof is. A wide spread usually points to tether routing or water access, not to robot performance.
Re-measure after any layout change. Moving the water point or adding an anchor can change the rate more than a different robot model would.
Key Takeaways
- Plan for 90-180 m2/h real-world, not the 300-400 m2/h spec claim.
- Heavy first-pass dirt cuts the rate by 40-55%.
- Tank refills, not battery, usually limit a cycle: 15-20 L covers 300-500 m2.
- Slow, lower-pressure passes protect coated glass and warranties.
- Never quote tender hours from spec-sheet speed; measure your own first pass.