Track four KPIs on a 2026 skylight cleaning robot – cleaned m2 per hour, litres per m2, streak reject rate and unplanned downtime hours – and you will know within a quarter whether the machine is paying off.
Vendors quote peak figures. Sites need honest monthly ones. The gap between the two is usually where the disappointment lives.
What is a realistic m2 per hour figure?
Ignore the marketing number and measure a full session, including setup and refills. Realistic 2026 figures on a cleanable flat or low-pitch roof: 250-450 m2 per hour on broad open glass, dropping to 120-220 m2 per hour where there are mullions, rooflights, vents or returns every few metres.
The obstacles are what kill throughput, not the glass. A roof that looks simple from below is often a maze of penetrations from above.
How much water should each square metre take?
- Straight-through: 0.7-1.2 litres per m2 on dusty glass.
- Recycling systems: 0.3-0.6 litres per m2 in practice.
- Anything above 1.5 litres per m2 signals a worn brush, a blocked nozzle or an operator compensating for poor technique.
Watch this KPI for drift. A slow rise over weeks is the earliest sign of a failing nozzle pattern, well before cleaning quality visibly drops.
How do you measure streak rate objectively?
Photograph the same six panels, from the same angle, at the same time of day, after each clean. Score each 1-5. This is crude but comparable, and it catches real drift. Two people comparing notes with no reference photos argue forever; a photo series ends the argument in a minute.
Reject above 15% of panels scoring 3 or lower and investigate – usually brush wear, water quality or a change in the water used.
What downtime should you accept?
Plan for under 5% unplanned downtime in year one and under 3% in year two, measured as hours lost against scheduled cleaning hours. If it exceeds 10%, stop troubleshooting and escalate to the supplier – that level usually points to a design or installation problem rather than routine wear.
Keep a simple failure log: date, symptom, part, hours to fix. Twelve months of that log is the most useful document you will have at the next budget review, and it is what suppliers respect when you raise a warranty claim.
Key Takeaways
- Expect 250-450 m2 per hour on open glass, 120-220 m2 per hour on obstacle-heavy roofs.
- Water use should sit at 0.3-1.2 litres per m2; a slow rise means a nozzle or brush problem.
- Use fixed-angle photo series to score streaks – aim for under 15% of panels rejected.
- Target under 5% unplanned downtime in year one and log every failure to support warranty claims.
How do you set baselines before you measure?
Run three cleaning sessions in the first fortnight and record everything: time on the roof, litres used, area logged, interventions, and photo scores. Average them. That is your baseline, and without it every later discussion is opinion against opinion.
- Time on the roof, door to door, not just machine-running time.
- Litres used, metered, not estimated from the tank gauge.
- Area covered per session, split by zone.
- Number of operator interventions.
- Photo scores for the six reference panels.
Baselines drift if they are not revisited. Re-measure each season, because glass conditions, brush wear and water quality all move the numbers.
What single KPI predicts trouble earliest?
Litres per m2. It rises before anything else: before visible streaks, before throughput drops, before a fault appears. A nozzle partly blocked puts more water on the glass and forces an extra pass, and the water figure registers it the same session. Watch that line on a weekly chart and you will catch most developing problems while they are still a cleaning task rather than a repair.
How often should KPIs be reviewed with the supplier?
Quarterly in year one, twice a year after that, with the failure log in hand. Bring the water and downtime figures; suppliers respond far better to data than to complaints. If the numbers cannot be explained, that meeting is usually where a warranty issue gets resolved before it becomes a formal claim.
How do you present KPI results to a building owner?
One page, three numbers, one chart. Show area cleaned against plan, litres per m2, and unplanned downtime. Owners do not want a telematics dashboard; they want to know the roof is being cleaned as promised and what it costs. The photo series can sit in an appendix for anyone who wants proof.
When a number is bad, say so before the owner finds it. A rising water figure explained in advance reads as competent management; the same figure discovered by a client reads as neglect. The reporting habit is the whole point of measuring anything at all.
What should trigger a full machine review?
Two consecutive months where water per m2 rises more than 15%, downtime above 10%, or photo scores falling below the baseline on half the reference panels. Any one of those means something has changed that routine maintenance has not fixed, and it is worth a half-day investigation rather than another round of cleaning.
Set the thresholds in writing when you set the baseline. Deciding them in the moment, under pressure to hit a schedule, is how a slow problem becomes an expensive one. The numbers only help if the reaction to them is agreed in advance.