Methodology

One source type only: the manufacturer. Product pages, downloadable spec sheets, manuals and official support documentation. Where a manufacturer states a figure in more than one place and the values disagree, we use the product page as the current claim and note the conflict where it matters.

We do not use retailer listings, marketplace bullet points, press aggregators or other comparison sites. Those chains introduce transcription errors that cannot be verified, and a wrong number presented confidently is worse than no number.

It means we looked and the manufacturer does not state that figure — not that the figure is unknown to the world, and not that we ran out of time. Every attribute in a category grid is checked against the model's own manufacturer sources before it is left blank.

We never estimate, interpolate from a sibling model, convert an unrelated unit, or fill a gap with a category average. In comparison tables, an attribute no manufacturer publishes for any of the selected robots is labelled as published by none, so you can see the gap is the market's and not ours.

Prices are manufacturer list prices, read from the manufacturer's own product page — from that page's structured product data where possible, with strict matching so an accessory or bundle price cannot be mistaken for the model's price. Each price shows the date it was last confirmed.

Checks run once a week, sequentially and slowly, with a capped number of pages per run. When a page cannot be fetched, or no price can be read confidently, we record the failed check with no price attached rather than re-stamping an old figure as freshly confirmed. A product page that returns a 404 flags the model for human review; it does not automatically flip availability to Discontinued.

The limits are worth stating plainly: this tracks list price, not the price you will pay. Retailer and marketplace pricing, sales, bundles and regional pricing are not tracked. Only models where we have a manufacturer product URL can be tracked at all. Every tracked price carries an inline chart of its recorded history, opened from the sparkline beside the figure.

Each guide declares the manufacturer-published metrics it ranks on — for example drawer capacity, supported cat weight, cleaning coverage, runtime, warranty length or price. Every qualifying model is scored on those metrics only, normalised across the candidate set, and ordered by the result. Models that do not publish the metrics a guide ranks on are placed last rather than being given an assumed value.

A model qualifies only if it is currently sold in the US market for that category and has enough published specifications to be compared at all. Nothing is paid, sponsored, or promoted. Commission relationships play no part in scoring.

RobotMatchup has not physically tested any product listed here, and no position in any list reflects hands-on use. There is no lab, no test protocol and no reviewer verdict on this site. What you get instead is the manufacturers' own claims, arranged so they can be compared honestly and so the silences are visible.

Data is revised whenever a manufacturer updates a published figure, and when readers send a source that contradicts ours. Corrections are made against the manufacturer source, not against opinion. Use the contact form and include the URL.