Read window cleaning robot reviews for four numbers: runtime at 30 °C, gasket replacement interval, low-battery behaviour, and spare part availability. A review that shows a streak-free pane but none of those four is a demo, not a review.
That is a hard position, and it rules out most of what is on YouTube.
What do most reviews actually show?
A unit running on clean, flat, room-temperature glass in a studio, with a wipe at the end. The wipe is the tell. If the glass starts clean and the pad removes a light film, almost nothing can fail.
Real use is different. A skylight in July has a 45 °C surface, a film of pollen and grit, and an edge with a drainage lip. The failure modes appear in the first ten minutes of that test, not in the studio.
Which numbers predict ownership cost in 2026?
| Data point | Why it matters | Red flag |
|---|---|---|
| Runtime at 30 °C | Heat derates cells and motor | Only 20 °C figures published |
| Gasket interval | Largest recurring cost | No stated interval |
| Low-battery behaviour | Decides callout risk | ‘Stops on glass’ |
| Spare availability | 2-week vs 3-month downtime | No parts list online |
Note what is missing. Suction force in Pascals. It sells machines and tells you very little about whether the machine will survive a year of dust.
Why do review scores cluster so tightly?
Partly genuine, partly structural. The units that reach reviewers are cherry-picked and tested for days, not months. Failure data lives with owners, not reviewers, and owners file it in a warranty claim rather than a video.
A better signal: look for repeated owner reports of the same fault. Three separate complaints about gasket hardening at 12 months is worth more than a five-star rating.
Complaints also cluster by climate. Coastal salt air attacks electronics faster than dry inland air. Middle East buyers report gasket issues sooner; northern European owners report mould in water tanks.
Which reviewer should you trust?
Someone who has used the machine for more than 100 hours, or who runs it as part of a service business. Short of that, prefer reviews that name the glass type and ambient temperature.
We keep our own notes on the skylight robot reviews archive, including units we have run past 1,000 hours. The Mamibot iGLASSBOT W120-P review shows the format: what the machine does, where it stalls, what it costs to keep running.
What is the biggest review blind spot?
Installation friction. Almost every review starts after the machine is on the glass. In practice, the hardest hour is getting it there: anchor fitting, cable routing, power access, and a ladder or hatch to reach the panel.
A machine that is 20% slower but installs in five minutes beats a faster one that takes twenty. Reviews rarely mention this because the reviewer has a fixed setup.
How should a buyer use review data?
- Compare like with like: same glass type, similar ambient temperature.
- Discount any runtime figure without a temperature attached.
- Ask the seller for the gasket part number and price before paying.
- Look for one negative review that names a specific failure, not a general complaint.
- Check whether the review covers setup, not only cleaning.
Finally, be suspicious of a review that praises everything. A machine with no stated weakness has either been used for an hour or written by someone with an incentive. Every unit we have handled has one clear weakness, usually the gasket or the cable reel.
Key Takeaways
- Four numbers decide ownership cost: runtime at 30 °C, gasket interval, low-battery behaviour, parts availability.
- Suction force in Pascals is a sales figure, not a durability figure.
- Repeated owner complaints beat star ratings.
- Installation friction is the most underreported part of any review.