How we source and present specifications
Sources
- Primary sources first. Values come from the manufacturer's own product pages, spec sheets, manuals, official stores, press releases, investor or regulatory filings, and official channels.
- Secondary sources are labelled. When a maker hasn't published a figure but a credible outlet has, we may show it labelled Reported, with its source; it counts for half in the rankings. Leaks and single-source claims are labelled ⚠ Unconfirmed and count for nothing. A robot's status (on sale, in a pilot) is only ever set from the maker or a named customer, never from a retailer listing or news report.
- Every value carries its source link, the date we checked it, and a confidence label.
Confidence labels
| Label | Meaning |
|---|---|
| Manufacturer | Stated by the maker in a primary source |
| Verified | Confirmed independently of the maker (our own measurement or an accredited test report) |
| Estimate | Calculated by us from official figures, e.g. battery Wh from mAh × volts. The method is shown |
| Reported | From a credible media outlet, retailer or trade show; not confirmed by the maker |
| ⚠ Unconfirmed | A leak, rumour or single source. Treat with caution |
| Not disclosed | The maker hasn't published it. We don't guess |
Most of our figures are the makers' own; the current share is shown on the Robot 500 page.
Definitions that matter
- Status: Concept (announced, no working unit shown) · Prototype (working units shown, no external customers) · Pilot (operating at a named customer site) · Pre-order (orders or deposits accepted, not yet delivered) · Shipping (delivered or generally orderable) · Discontinued.
- Degrees of freedom: independently actuated joints, as the maker counts them. We note the counting method where it is known.
- Payload and runtime: shown with the conditions the maker gives (e.g. per arm, or while walking). Where the maker gives none, we say so.
- Suction (Pa): shown as each maker states it. There is no common test standard, so we never rank vacuums on Pa alone.
- Mower area: shown with how the maker rates it (maximum lawn size, per charge, or per hour).
- Ranges stay ranges; we never collapse "75–85 kg" into a single number.
- Conflicts: if a maker's own pages disagree, we show the most recent figure and note the other.
Quality control
Each record is built by one research pass and re-checked value by value against its sources by a separate quality-control pass before publication. We track our error rate, and every correction is logged publicly in our corrections log. What the data and rankings can't tell you yet is listed on our known limitations page.
Comparisons and rankings
We create comparison pages only for robots people genuinely weigh against each other, not every possible pairing.
Our robot rankings, the Robot 500, are computed only from the sourced data on this site, using a published method: robots are compared only with others of the same kind, on sourced performance (70%) and real-world proof such as sales, deployments and certifications (30%). Each rank carries an evidence grade, and robots with too little evidence are listed on a watchlist, not ranked. The full formula, weights and version history are on the Robot 500 method section. No company can pay for a better position, there are no manual overrides, and the criteria change at most once a year, announced in advance. Sponsorship, if we ever accept it, will never affect results.