How we check what we publish

Where the numbers come from

A price is read from the vendor’s own pricing page — never from another comparison site, never from a press release, never from memory. Each figure carries the date it was read and links to the page it was read from.

Vendors show different prices depending on where you connect from. Not converted prices: different ones — we measured Capsule showing €15, €28 and €49 from Germany and US$18, US$36 and US$54 from the United States. Each figure records its currency and pricing region, every page declares its reading conditions next to its dates, and our collector’s own vantage is measured, not assumed: since 23 August 2026 it reads from a datacenter connection in Germany, in English, and a vendor that prices by visitor region is published as a fact about that vendor.

How a figure stays verifiable months later

A link and a date say where and when. They do not say what the page showed. In three months that URL will show a different figure, and our history would become a claim we make about ourselves.

So we keep the page text we read, stored under the SHA-256 hash of that text. Anyone can check that a stored reading is the one we used:

sha256sum <file> # equals the file name

The file contains the text and nothing else — no header you would have to take on trust. The URL, date and region live in an append-only log next to the hash.

What we do when we cannot check

Some vendors block automated checks, and some price tiers only appear once you declare a contact volume we have not set. Where that happens the page says so, names the vendor, and shows nothing rather than a number nobody has verified. A field we could not qualify is marked as not enough evidence to publish, with the reason.

A reading taken by hand expires after 30 days. After that the page reports the price as not enough evidence to publish rather than showing a figure that has aged out.

What we do not publish

• No score out of ten, no star rating. A number without a published method is an opinion wearing a number.
• No rating imported from a review platform. That is someone else’s method, and we cannot check it. What reviews complain about is a fact with a source and a date, and that we can publish.
• No promotional price as the comparison price. Several vendors advertise an introductory rate; on one of them it is a tenth of the durable price. We publish what you pay once the offer ends, and name the promotion and its window next to it.
• No product experience we have not had. We can tell you what we read and what happened while reading it. We cannot tell you what a tool feels like after three months, because we have not used it for three months.

How the recommendation is produced

Criteria, weighted differently for different buyers. We publish the winner, the runner-up, and the gap between them on a zero-to-one scale — including when that gap is zero, which means the two are level and we say so instead of picking one.

Some differences do not enter the ranking because we cannot measure them from outside — the depth of a link index, for instance. Where one of those decides a purchase, we say it in words and label it as unmeasured, rather than turning it into a criterion that would conveniently break a tie.

How this page makes money

Some outbound links are affiliate links. Where a page carries one, a disclosure appears at the top of that page, and it is computed from the links themselves rather than written by hand — so it is true on the day you read it. Recommendations are decided before monetisation is considered, and a product with no affiliate programme can win.