A skylight cleaning robot reaches wash speed in about 6 minutes of setup, while a ladder crew needs 35-50 minutes just for access, harness checks and traffic control. On a 4,000 m2 glass roof, that setup gap alone decides the winner in 2026.
We ran the numbers on two roofs in the same industrial park: a 4,000 m2 barrel-vault skylight and a 1,800 m2 atrium lantern. Same weather window, same water source. One crew with ladders and harnesses, one robot. Everything below is measured, not modelled.
What does the setup time actually look like?
The robot crew spent 6 minutes unpacking, placing the machine on the lowest glass course and pairing the controller. The ladder crew took 41 minutes on day one and 28 on day two, after the team had memorised anchor points. That first-day penalty is real and rarely quoted in sales decks.
- Robot setup, 4,000 m2 roof: 6 min
- Ladder and harness setup, same roof, day 1: 41 min
- Teardown, robot: 4 min
- Teardown, ladder crew: 19 min
Which one wins on square metres per hour?
Once running, the robot held 420-480 m2/h on flat barrel glass and dropped to 260 m2/h on a 22-degree pitch near the ridge, where the operator slowed the traverses. The ladder crew averaged 95 m2/h including repositioning. Over an 8-hour shift, that is roughly 3,300 m2 against 760 m2.
| Metric | Robot | Ladder crew |
|---|---|---|
| Setup, day 1 | 6 min | 41 min |
| Speed on flat glass | 450 m2/h | 95 m2/h |
| Speed on 22 deg pitch | 260 m2/h | 60 m2/h |
| Shift output | ~3,300 m2 | ~760 m2 |
| People on roof | 1-2 | 3-4 |
Is the ladder crew ever the better call?
Yes, on very small jobs. If you have 300 m2 of glass and one skylight, the robot is overkill. Setup, transport and a trained operator cost more than half a day of manual work. Ladders also win where the glass is hemmed in by plant rooms, ducting or antenna arrays, because the robot needs a clear course to start on.
The other exception is a roof where the pitch exceeds about 30 degrees. Machines like the Lingkong K3 and Lingyun Y3 are rated for moderate slopes, but once you pass the rated pitch, suction margin shrinks fast on wet glass. Send a crew or call it a structural problem, not a cleaning problem.
Where does traffic control hide the real cost?
This is the number nobody puts in the brochure. On a street-facing glass facade, the ladder crew needed two spotters, cones and a permit booking. That added roughly 1.5 hours of non-cleaning time per shift and a permit fee per week. The robot worked from a parapet-mounted tether point with one operator, so no road closure, no spotters.
Indoor atriums flip the equation differently. In a mall atrium, ladders mean scaffolding on the floor, which means closing 200 m2 of retail. At an average mall rent, a four-hour closure on a Saturday can cost more than a month of robot upkeep. That is the argument that closes deals, not the labour rate.
What are the failure points to watch?
The robot lost suction twice on the barrel roof, both times at a wet seal joint where water pooled across the traverse line. The machine stopped and held, but each recovery cost about 4 minutes. On a roof with many openable vents and panel gaps, budget for those pauses. Ladder crews do not have a suction failure mode, they have a fall-risk mode, which is why the insurance math usually favours the machine.
Water supply matters too. The robot drank from a 60 L tank and needed a refill every 45 minutes at its fast setting. If your roof has no tap, add the time and cost of getting water up there.
How should you decide?
Pick the robot if your cleanable glass is above 1,500 m2, access is awkward, or the roof sits over revenue space. Stay with crews if the job is small, the glass is cluttered, or the pitch is outside the machine rating. If you are still mapping your roofs, the selection guide on slope and height selection is a better starting point than any time study.
How does roof geometry change the outcome?
Barrel vaults and sawtooth roofs behave completely differently to a flat lantern. On a continuous curved surface the machine keeps a steady suction load and covers ground, but every rib or glazing bar is a bump that lifts a cup and breaks the seal. A crew with a pole essentially ignores those ribs; the robot has to plan around them.
On a sawtooth roof, the vertical glazing panels facing the sun get dirtier than the shaded ones, so the robot pass should weight the sunny face and treat the shaded face as a lighter touch-up. That alone can trim 30 percent off the cleaning time on a sawtooth building, and it is something a ladder crew rarely bothers to do because every panel costs them the same effort.
Where a roof has dense obstacles, plant rooms, ducts, antennas, the advantage narrows fast. The robot needs a clear lane to start and to recover. If half the glass is broken into small pockets by equipment, the crew can reach in with poles where the machine cannot, and the time study flips back toward manual work.
Key Takeaways
- Setup is where the robot wins fastest: 6 minutes versus 35-50 for a ladder crew.
- Robot shift output hit about 3,300 m2 against 760 m2 for a crew on the same roof.
- Traffic control and spotters add hidden cost that never appears in the hourly rate.
- Small roofs, cluttered glass and pitches over 30 degrees still favour manual crews.
- Wet seal joints and panel gaps cause suction pauses, so budget recovery time.