Test Lab · Approach
How I Research and Test Robot Vacuums
Most of what I publish here is research. Bench work is planned for when a unit is available. This page explains the difference - and the rules I hold both to.
How I work
This site answers one question: will this robot actually work in your home? I answer it in two ways, and every page here states clearly which one it is.
Research-based content
Most of the articles, guides and vs. comparisons on this site are built from public information: manufacturer specifications, official documentation, verified prices and published testing I can link to. I compare the numbers that decide real use - clearance and dock dimensions, battery life, filter and part costs, features - and I link the sources for every claim that matters.
These pages are honest about what they are. They tell you what the specifications say, where prices stand today, and which questions you should check on the unit itself before buying. A research page never pretends to be a lab test, and it does not carry a “Tested” badge.
Bench tests
If I physically have a unit - bought at retail, loaned by a brand, or sourced any other way - it will go through one fixed protocol instead of anecdotal impressions. Same rigs, same debris charges, same room, same tester, so future results are comparable across models.
The protocol covers the Home Fit measurements that decide whether a robot works in your home:
- Threshold climbing on a 10-90 mm ramp
- Furniture clearance in 1 mm steps, entry and escape measured separately
- Dock footprint including the clearance it needs to seat
- Pickup on hard floor and carpet with weighed debris charges
- Hair handling, separated from pickup
- Rugs and fringe: black surface, ingestion, displacement
- Navigation and obstacles: cables, socks, chair legs, doorways
- Noise at a fixed distance in every cleaning mode
- Privacy and cloud behaviour, tested with the internet blocked
- Running cost from verified consumable prices
- Long-term reliability tracked over months where possible
If a bench run is published, every figure will carry the firmware it was measured on. Failures will be recorded and shown, not edited out: stuck events, dock clogs and app faults are results too.
How content is labelled
- Tested - the unit went through the full protocol above on my bench.
- Research-based - the page is built from specifications, verified prices and linked sources, and says so at the top.
- Not tested yet - no bench data exists for this model. A blank cell means not tested; it never means zero, and it never means fine.
Principles
- Every number is labelled for what it is. Data from a spec sheet says it is from a spec sheet. Any future bench data will say it is from my bench.
- Every future bench test uses the same protocol. A result that cannot be compared is not worth publishing.
- Every source is linked. Claims built on research point to where they came from, with the date where prices are involved.
- Firmware changes results. Any figures measured on a future bench run will carry the firmware they were measured on, and material updates trigger a retest.
The rules that keep this honest
Brands cannot approve, preview or influence anything here - a written policy, not a sentiment. See editorial independence. How units are sourced and labelled is documented in the unit source policy. When a result is wrong, I correct it in place, date the correction and keep the original figure visible - see the corrections and update policy.
The person behind all of this is Christian Huber, a one-person operation based in Germany.