How we analyse and score
We compare window-cleaning robots using their technical specifications, manuals and publicly available buyer reviews. Our analysis covers supported surfaces, minimum dimensions, movement, spraying and safety features in particular.
What we mean by “analysis”
We do not physically test products. Advertised performance is attributed to manufacturers, and buyer feedback is presented as personal accounts. Any specification missing from the sources or contradicted by them is flagged as uncertain.
A manufacturer's claimed specification is not measured performance, and a customer review is not a test result. We keep these three types of evidence separate, both on the page and in our calculations.
Our evidence trail
- 1Exact modelThe full model name, variant and market. We never combine data from two versions of the same product.
- 2Published dataManufacturer specifications and retailer listings for the exact model, recorded on a specific date.
- 3Comparison across our catalogueEach value is compared with every other model on the same metric.
- 4Verdict in contextWho the model suits, and when to choose another.
Missing data stays missing. We never guess it, replace it with an average or count it as zero.
How to compare window-cleaning robots
These criteria are fixed across the category and applied in the same order to every model. This lets you compare two analyses line by line, instead of reading two reviews that cover different things.
- Suction pressure higher is better · pa
- Water tank capacity higher is better · ml
- Backup battery holding time higher is better · min
- Equipmentdocumented features tracked for this category: the more there are, the higher the score
A criterion is only scored if its specification is documented for the model. Across the site's 27 scored analyses: Suction pressure out of 14, Water tank capacity out of 18, Backup battery holding time out of 16, Equipment out of 27.
That's why each product page shows between 1 and 4 scores. Under “Our scores explained”, each page lists any criteria left unscored and explains why.
A model is not penalised for a feature it never promised. It is assessed in the context of its price and intended users, using criteria relevant to its category.
How the score is calculated
The score out of 10 is the average of individual scores, one for each criterion in the table above. Each score reflects more than an impression: it shows themodel's ranking across the entire catalogueon that metric. That's why each row includes supporting figures, such as “210 min, ahead of 80% of models in the catalogue”. You can cross-check them.
When our data does not cover a criterion for a particular model, the row makes this clear: it is presented as anassessment based on published specifications, rather than a measurement. The page clearly distinguishes between the two.
What we never do:
- fill in a missing specification with a guess;
- score two models using different approaches;
- declare an overall winner based on results that depend on user needs;
- publish a score when there is too little data to support it.
A score reflects our assessment at a given time. It is relative to the catalogue: whenever we update scores, every product is reassessed against the current catalogue, so scores can change as new models are added. Scores last updated: November 1, 2026.
A real example: Ecovacs WINBOT MINI
Here is the calculation for the “Water tank capacity” score for Ecovacs WINBOT MINI, exactly as shown on its product page, one of the best-documented analyses on the site.
Water tank capacity: 7.9/10
Two model counts: the rank and average include product pages on this site scored for the criterion; the calculation compares the model with all catalogue models whose specification is verified, with or without a product page. A score assigned at publication may remain without a verified specification, so the first count can sometimes be higher.
Ecovacs WINBOT MINI scores 7.9/10 on this criterion. Across the 18 product pages on the site with a score for this criterion, the average is 8.1: below average, Rank 7 out of 18.
| Shared baseline: last place in the ranking | 6.5 |
|---|---|
| Water tank capacity: 60 ml ; 4 perform worse, 6 perform better, 8 are tied (including this model), among the 18 catalogue models with this specification verified | +1.4 |
| Water tank capacity score | 7.9/10 |
Calculation: (4 + 8 ÷ 2) ÷ 18 = 44.44% of the ranking; 6.5 + 44.44% × 3.1 = 7.9/10, rounded to one decimal place.
Complete or partial documentation
Next to the overall score, each product page shows whether the category's 5 key specifications are confirmed by a verified source (an archived retailer page or manufacturer documentation): Suction pressure, Water tank capacity, Backup battery holding time, Power supply during cleaning, Manufacturer's intended use. It shows “Documentation: complete” when all are confirmed, or “partial” otherwise, with missing details shown on hover. In the specifications table, any of these 5 specifications that are not confirmed are marked “unverified”; other rows in the table are not covered by this check.
Partial documentation does not lower the score: a criterion with an unconfirmed value is not scored and is excluded from the average. The only exception is a score assigned when the analysis was published, which is retained while the product page has fewer than 3 verified criteria; the page's “Our scores explained” section highlights this.
Currently, 6 product pages have complete documentation and 21 have partial documentation.
Our sources and how we weigh them
- Manufacturer documentationOfficial pages, manuals and specification sheets. We use these to identify the manufacturer's claims. Those claims are always presented as such.
- Retailer listingsThese confirm the exact model, included accessories and recorded price for the target market. A limited-time offer is never presented as a permanent price.
- Customer feedbackWe look for recurring themes, rather than isolated comments, taking into account reviews covering different variants and the sample size.
- Published specialist analysesWe use these to check claims against observations. We don't copy their verdicts or average their scores.
When sources disagree, we highlight this rather than force a conclusion.
Prices
We don't display prices next to retailer links. A price checked once soon becomes outdated, and an incorrect figure next to a buy button lets down the reader it was meant to help. Prices shown in our comparison tables and filters help place a model within its range and let you sort by budget: the price that applies is the one displayed by the retailer when you buy.
Independence and affiliate links
No brand pays to feature in our picks or improve its score. Some retailer links are affiliate links: if you buy through them, we may earn a commission,at no extra cost to you.
Whether we earn a commission, and how much, plays no part in our criteria. Affiliate income funds the site; it does not determine our rankings.
Updates and corrections
We update a review when the model changes, when information proves incorrect, or when new models change its standing in our catalogue. The last updated date is shown on every page.
This site's content is written and checked by Romain Brunel. Spotted an error, outdated information or a missing model? Let us know and we'll update the page.
Read our reviews →