Methodology

Source authority is decided per field, not globally. A manufacturer’s store page is the best source for what a robot costs today and a poor source for its rated payload, where the technical manual wins. Applying one hierarchy to every field is how aggregators end up publishing a specification the manufacturer’s own manual contradicts.

Price and availability

1. Manufacturer store or checkout page
2. Manufacturer product page
3. Authorized distributor
4. Structured product feed
5. Retailer, then marketplace
6. News reporting, only where nothing above exists

Technical specifications

1. Manufacturer technical documentation and manuals
2. Manufacturer developer documentation
3. Manufacturer product page, then store page
4. Authorized distributor
5. News, then independent community testing

Developer capability and measured performance

Developer fields are resolved from developer documentation first, because a product page describes what a robot does and a developer portal describes what you may build on it. For fields that explicitly measure observed behaviour — battery life under load, for instance — independent testing outranks the manufacturer, and the measured figure is stored beside the rated figure rather than replacing it.

When two sources of the same rank disagree and the field allows only one value, the conflict is recorded and the record is held for review instead of publishing a guess.

Every price, specification and availability status carries the date it was observed, and that date is shown wherever the value is shown. A record is rechecked on a 120-day cycle; past that window it is marked as due for a recheck rather than quietly presented as current.

A robot reaches a public page only after a human has reviewed the evidence behind its price status, availability and at least 5 specifications. An official source nobody has opened is still unverified, and it does not satisfy the gate.

Verified

Stated by the manufacturer

Confirmed

Stated by an authorized seller

Reported

Reported by a secondary source

Estimated

Calculated by RobotCosts

Starting prices

A “from” price is recorded as a starting price and labelled as one everywhere it appears. It is never presented as the price of the robot as configured in the photograph, and it is never silently upgraded to a list price.

Shipping, tax and duties

Whether a price includes shipping or tax is recorded as stated by the seller and shown as a qualification beneath the figure. Where the seller does not say, RobotCosts does not assume: the qualification is absent rather than invented. Import duties are never included, because they depend on a destination the site does not know.

Currency

Prices are stored in the currency the seller quoted. Figures in different currencies are never averaged and never converted for the purpose of a statistic, because a conversion introduces a rate and a date that were not observed. A category distribution is computed within a single currency and says so.

Quote-only and undisclosed prices

Most commercial robotics is sold by quote. When there is no public price, the page says “Quote required” or “No public price” as the dominant answer. It never shows a zero, a blank state, or an estimate dressed as an observation. A quote-only robot is not a failure of the database; it is an accurate description of how that robot is sold.

Lease, subscription and used prices

Lease, rental, subscription and used prices are recorded and shown, but they are kept distinct from a one-time purchase price and are excluded from purchase-price statistics. A monthly figure and an outright price are not the same number.

A category distribution uses one representative price per model, so a manufacturer shipping twenty variants cannot move a median. Distributions are published only at 10 or more compatible prices; below that the page describes the inventory instead. Aggregate pages become indexable at 2 models for a manufacturer and 8 for a category.

• No ratings, reviews or stars. RobotCosts does not collect customer reviews, so it does not display them and does not emit rating structured data. A five-star average nobody submitted is a fabrication.
• No invented prose. The editorial block on a robot page restates fields the page already shows with sources attached. Where the evidence is thin, the block is short or absent.
• No interpolated history. Price history is a list of dated readings. No line is drawn between two observations, because the days in between were not observed.
• No merchant claims. RobotCosts is an editorial aggregator, not a seller. It never claims merchant-listing eligibility for products it does not sell.

Every robot page carries a correction form. A report identifies the field, what the correct value is and where it can be verified. Corrections that come with a source are actioned first, because they can be checked. Accepted corrections update the record and appear in that robot’s change log with the date.

Manufacturers and distributors who want a price updated should use the same form; there is no separate channel, and no commercial relationship changes how a correction is handled.

Images are used only where the rights allow it, with credit and attribution recorded alongside the asset and shown on the page. Where no cleared image exists, the page uses a neutral placeholder rather than a photograph taken from a source that did not grant permission.

Some outbound seller links may carry affiliate tracking. Those links are marked in the interface and tagged rel="sponsored nofollow" in the markup.

Commission never influences which offer is shown as the headline price. Offer ranking is decided by evidence quality, seller authorization and whether the item can actually be ordered — in that order, implemented in one place in the codebase, with no input for commercial terms. If an affiliate link and a better-evidenced non-affiliate offer disagree, the better-evidenced one wins.

See also the source policy and about RobotCosts.