Managing eight skylight cleaning robots is an operations job, not a maintenance job. At that scale you need three things: a rolling 30-day schedule, a battery and water rotation plan, and a parts buffer covering roughly 15 percent of the fleet. Sites that skip the schedule plan lose 20 to 30 percent of fleet hours to avoidable idle time.
Below two machines, you can run on memory. Above four, memory fails you. The failure is rarely dramatic; it is one robot sitting idle for a week with a flat battery while another runs double shifts.
What breaks first when a fleet grows?
Water logistics, every time. One robot needs one tap. Eight robots across four buildings need a refill routine that does not collide with building hours, tenant access, or cleaning staff breaks. If the machines use tether feed, you also inherit hose routes and trip hazards.
- Battery rotation: stagger charge cycles so no two machines sit idle at once.
- Consumable carts: one mobile cart per two machines keeps cups, blades, and filters within reach.
- Scheduling: a rolling 30-day board beats a spreadsheet nobody updates.
- Spares: buffer 15 percent of the fleet in rollers and squeegee blades.
Which data is worth collecting?
Not everything. Four fields earn their place: machine hours, area cleaned, fault codes, and consumable replacements. Anything else is noise unless you have a specific problem to chase.
| Metric | Why it matters | Review cadence |
|---|---|---|
| Machine hours | Drives planned maintenance | Weekly |
| Area cleaned | Validates productivity | Monthly |
| Fault codes | Early warning of suction or drive issues | On event |
| Consumables used | Forecasts reorder timing | Monthly |
Fleet-tracking setups like those used around the Lingkong K3 report these four automatically. The value is not the dashboard itself; it is having the numbers in one place when a machine starts slowing down.
Should you centralise or distribute machines?
Distribute coverage, centralise maintenance. Keep machines near the roofs they clean, but run one workshop for deep servicing. If each site services its own units, standardisation collapses, and you end up with eight machines in eight conditions. Rotate every unit through the same workshop on the same interval, and compare their fault logs side by side.
Who should not run a fleet?
A single building with one roof and one operator. A fleet is overhead you will not recover. If your total glass area is under 3,000 m2 and you clean fewer than six times a year, two machines is already generous. We have seen facility teams buy four units for a roof that needed one, then complain about maintenance load. Match the fleet to the workload, not the budget. Our earlier eight-machine notes dig into the rotation maths, and the contact page covers fleet sizing.
Key Takeaways
- Past four machines, a rolling 30-day schedule replaces memory-based planning.
- Water logistics is the first thing to break as a fleet grows.
- Buffer about 15 percent of the fleet in rollers and squeegee blades.
- Track four fields only: hours, area, faults, consumables.
- Distribute coverage but centralise servicing, or units drift into different conditions.