Skylight cleaning in 2026 costs roughly 3-9 USD per m² per visit with rope access or a lift, and 2,000-8,000 USD a year on a maintenance contract. A dedicated skylight robot runs 6,000-25,000 USD once. Break-even lands at 1-3 years for buildings with more than about 200 m² of glass roof.
Those are wide bands because the job is not one job. A 60 m² atrium skylight over a shopping centre escalator is a very different problem from 800 m² of factory roof glazing. Here is how the numbers actually break, and where quotes go wrong.
What drives the per-visit price?
Access, mostly. The cleaning itself is maybe 40% of a rope access invoice. The rest is rigging, permits, a spotter, an exclusion zone and the two hours before anyone touches glass.
- Covered atrium, interior access: 3-5 USD/m², no rigging, often done at night
- Low-slope roof glazing, ladder or platform: 4-7 USD/m² plus equipment hire
- Steep or fragile glass roof, rope access: 7-12 USD/m², and some buildings get refused outright
- Emergency or one-off callout: add 30-50% for mobilisation under a week
Fragile overhead glazing is the killer. Laminated panes that cannot take foot traffic force rope access technicians to work from above on a suspended platform, and the rate climbs with every metre of drop below them. Buildings where the glazing is over an occupied space also carry a debris-netting cost that never appears in the headline rate.
How often does a skylight actually need cleaning?
Twice a year is the default in temperate Europe and North America. In dusty or pollen-heavy regions, or under flight paths, four visits is realistic. Coastal sites get salt film on top of dust and often need quarterly work just to keep the light transmission above 70%.
And that last number is the one that gets facilities managers to sign. A dirty glass roof can lose 15-30% of daylight transmission. For a building relying on daylight to cut lighting load, that is a measurable energy line every month, not a cosmetic one. Dust plus bird droppings on a northwest-facing skylight is worse than dust alone because the deposit holds moisture.
When does buying a robot beat the contract?
Run the arithmetic on annual spend. A 400 m² atrium cleaned twice a year at 5 USD/m² is 4,000 USD annually. A 600 m² factory roof glazing cleaned quarterly at 6 USD/m² is 14,400 USD. Over five years that second case pays 72,000 USD in access invoices and still has no asset at the end of it.
| Glass roof area | Contract, 2 visits/yr | Contract, 4 visits/yr | Robot payback |
|---|---|---|---|
| 150 m² | ~1,500 USD | ~3,000 USD | 5+ years, usually not worth it |
| 400 m² | ~4,000 USD | ~8,000 USD | 2-3 years |
| 800 m² | ~8,000 USD | ~16,000 USD | 1-2 years |
The payback column assumes a 10,000-20,000 USD machine, two operators trained, and a realistic 80-120 m²/h on a stepped glass roof. That rate is the fragile assumption. Bars, ridge lights and ridge caps cut it hard, and a roof designed without robot access in mind can halve it.
What costs do the brochures leave out?
Water. A roof-class machine moves a lot of it, and if you are not near a riser you are carrying 20 L canisters up a ladder. Some sites pay for a temporary tank and pump, which is a capital line nobody plans.
Then training. Two people need to know how to rig, run and recover the machine, and that knowledge leaves with them when they resign. Budget one refresher a year. Also budget the fall-arrest kit for the operator who walks the roof to reposition the robot: that person is at height too, and the machine does not remove that exposure, it just shortens it.
Last one, and the most expensive: the day the robot cannot climb a particular bay. Steep north lights, a deep gutter, a ridge detail designed in 2004. You will still call a rope team for that 8% of the roof, and the contract price for a small awkward area per visit is worse than the full-roof rate per m².
Who should keep paying the contractor?
Buildings under roughly 200 m² of roof glazing. Museums and heritage structures where the glazing itself is fragile or listed. Landlords who want the invoice, the liability and the insurance bundled into one vendor and are happy to pay for that. All three are legitimate reasons, and none of them are about the robot being bad.
If you are in the buy camp, compare the hardware honestly before committing. Our skylight robot reviews index covers grip, climb rate and real running cost model by model, and the Mamibot iGLASSBOT W200 SkyBot review is a fair read on where a prosumer unit fits between a consumer suction robot and a facade system.
Is the ROI case different for a rented building?
Yes, and it usually kills it. A capital purchase that pays back in three years does not work on a two-year lease with a service charge already contracted to a cleaning firm. In that case the better lever is renegotiating the schedule: move from four visits to two and accept a slightly duller roof, then spend the difference on a water filtration upgrade that keeps the glazing clear for longer. Cheap, boring, effective.
What about a panel array that was never designed for robots?
Most retrofit arrays are not. The gaps between rows are often too narrow for a tracked chassis, the DC cabling runs across the frame, and the mounting clamps sit proud by 30-40 mm. A robot that crosses 20 mm glazing bars will jam on a 35 mm clamp. Before pricing any machine, measure the row gap, the clamp height and the walkway width, then check those three numbers against the datasheet. Two out of three arrays we look at fail on the row gap alone.
Key Takeaways
- Rope access or lift cleaning runs 3-12 USD/m² per visit in 2026, with access as the main cost.
- A dirty glass roof loses 15-30% of daylight transmission, which shows up on the energy bill.
- Break-even on a 10,000-20,000 USD robot arrives at 1-3 years above roughly 200 m² of glazing.
- Small, fragile or heritage glazing: stay on a contract. Large regular roofs: the machine wins.