Running eight skylight cleaning robots is a different job from running one. In 2026, the fleet that works has a shared spares pool, a fixed cleaning calendar by roof, and two trained operators per machine, not one.
One robot is a tool. Eight is a small operation with logistics, and the failure modes shift from the machine to the schedule. Below is what we learned sizing a fleet across a mixed portfolio of malls, offices and one terminal.
How many operators do you really need?
Two per machine, minimum. Not because a machine needs two people, but because holidays, sickness and shift patterns will otherwise strand a robot. An idle machine still costs money in depreciation and storage. On an eight-machine fleet, that is 16 trained operators, or a smaller core of 10 cross-trained across three buildings plus a relief list.
- Two named operators per machine.
- One relief operator per two machines for absence cover.
- One technician who owns spares and consumables across the fleet.
What spares should you hold?
Suction cups and pad sets wear fastest. On a mid-use fleet, cups last months but a single failed cup can stop a clean, so keep a small stock. Brushes, squeegee blades and water filters are the other consumables. Beyond that, hold one spare battery per two machines and a spare controller per four. Motors and fans you order on failure, not on stock.
The headline number to watch: a fleet that shares a single spares pool runs lower total stock than eight buildings each holding their own. Consolidate.
Does software actually help a fleet?
Partly. Scheduling and log capture are genuinely useful: knowing which roof was cleaned, when and by whom turns a compliance question into a lookup. Predictive maintenance claims are more marketing than substance at this scale, so do not buy a platform for that alone. What does help is a simple shared calendar and a photo log per clean, which many teams already have without extra software.
How do you schedule across a portfolio?
Fixed cycles beat reactive cleaning, but only if the cycle matches the roof. A mall atrium needs a 6-week cycle, a factory monitor roof 8-12 weeks, a terminal glass wall every 4-6 weeks. Group roofs by geography so one operator-day covers two buildings, and put the hardest roofs first in the morning while the operator is fresh. Our cleaning schedule guide has the cycle logic.
What is the biggest fleet risk?
Single-operator dependency. If one person holds the knowledge for four machines and leaves, you lose weeks. Write down the method statements, keep them in a shared folder, and rotate operators so the knowledge does not sit in one head. The second risk is water logistics: eight machines in one portfolio need reliable water points, and a site without one quietly wastes an hour per clean.
Who should not build a fleet?
Anyone with fewer than about 15,000 m2 of cleanable glass total, or roofs spread too far to share operator time. Below that, one machine plus a good rental partner covers the peaks, and you avoid the staffing overhead entirely. Scale only pays when the roofs are close enough to share a person.
How do you keep utilisation high?
Utilisation is the number that decides whether a fleet pays. Track machine-days worked against machine-days available each quarter. A fleet below 50 percent utilisation has too many machines or too few roofs. The fix is usually geographical grouping: move a machine to a cluster of buildings within an hour’s drive rather than leaving it idle near one large site.
Rotate machines across sites so wear spreads evenly. One machine on the hardest roof for two years ages far faster than the same hours spread across three roofs. Even rotation keeps the fleet’s average health up and avoids one chronic problem machine.
How should reports be structured?
Keep it boring. One page per building per cycle: date, operator, glass area cleaned, water source, any incident, after-photos. That page answers nearly every client question and most compliance requests. Fancy dashboards rarely change a decision; a consistent one-page log does.
What is the second biggest risk?
Consumable drift. If nobody owns the spares pool, pads and cups run low quietly and a clean is missed because a cup failed. Assign one technician as owner, and set a reorder trigger, for example two spare cup sets per machine on the shelf at all times. It is a small discipline that keeps a fleet moving.
How do you onboard a new building?
Add one building at a time. Survey the roof, note glass type and pitch, run a trial clean, agree the cycle, then enter it into the fleet calendar. Onboarding four sites at once overloads the operators and the spares pool, and small problems compound into missed cycles across the whole portfolio.
Pair a new building with an experienced operator on the first two cycles, then hand it to the site team. That is how the fleet keeps quality steady as it grows.
A final note on cost: a fleet’s per-machine cost falls as it grows, but only to a point. Past eight to ten machines, you need a coordinator, and that fixed cost is real. Size the fleet to the roofs you have, not to the roofs you hope to win.
Key Takeaways
- Two trained operators per machine plus relief cover is the working minimum.
- Share one spares pool across the fleet instead of stocking per building.
- Cycles differ by roof: 4-6 weeks for terminals, 8-12 for factory monitors.
- Group roofs geographically so one operator-day covers two buildings.
- Under about 15,000 m2 total, one machine plus rental beats a fleet.