Every field we collect about a business, its source, and the basis we hold it on
| Field | Where it comes from | Basis |
|---|---|---|
| Domain | Where it comes fromCertificate Transparency logs. A certificate has to be published to one of these before a browser will trust it, and the logs are public and append only. We take only some address endings: .com, .org, .net, .info, .ai, .app, .dev, .online, .store, .site, .pro, .tech, .live, .studio, .space, .uk, .ca, .in, .ch and .nz. Every other ending is passed over before the site is ever read, because the registry behind it does not tell us when the website went live, and a lead we cannot date is one we do not show. | BasisPublic record |
| Business name, trade, platform, tech stack | Where it comes fromRead from the homepage the same way a browser reads it. If the homepage lists no way to reach the business we try a contact or about page, and stop at the first one that does. | BasisPublished by the business |
| Country | Where it comes fromThe ending of the website address when that ending belongs to a country, which answers most often. Otherwise what the site publishes about itself: an address in its own markup or written on the page, a phone number, money only one country uses, the language the page declares. Left blank when none of those is reliable. | BasisRegistry record, then published by the business |
| Town and state | Where it comes fromOnly an address the business publishes in its own markup. Never guessed from a phone code or a place name in the page text, so most rows carry neither. | BasisPublished by the business |
| Contact address | Where it comes fromOnly what is printed on the business's own site, and only once its own mail server has said the mailbox exists. An address it will not confirm is dropped. The default is a role address such as info@ or hello@. | BasisPublished by the business |
| Social links | Where it comes fromThe profiles the site links to itself, on the handful of networks we look for. A link to any other network is not stored. | BasisPublished by the business |
| What the business already runs | Where it comes fromWorked out from the same fetch, by reading what the page loads and links to. It covers the marketing and analytics tools on the site, the payment, review and booking services it uses, whether a phone number or WhatsApp is shown, and how the page itself is built. A buyer sees this as the business's stack before spending anything. | BasisRead from the page |
| What the site is missing | Where it comes fromThe gap list and its count, worked out from the same fetch as everything above: what a visitor would find absent, such as no way to book or a page that does not fit a phone. | BasisRead from the page |
| When it was first seen, and last read | Where it comes fromThe date the certificate turned up in the logs, and the date of the most recent fetch. | BasisPublic record, and our own run |
| When the website went live | Where it comes fromThe date the website first came online, taken from the public domain registry for the address a visitor actually lands on. It is not the date the business started trading. A shop that has traded for twenty years and only just built a website reads as a new website, because that is what it is. Where the registry does not answer, the entry says it does not know rather than guessing. | BasisPublic registry |
| The year the business gives for itself | Where it comes fromA year read off the business's own homepage: a copyright line, a founding date in its markup, or wording such as "since 1998". It is the business's own statement about itself, nothing is inferred from it, and it never decides whether a business is listed. | BasisRead from the page |
| How it got here, and whether it is in the feed | Where it comes fromThat the domain came from the certificate logs, and a state on the row. It reads qualified once the row has passed the filters. A new website is also put through the business check below before it is shown: the row reads pending while that check waits for an answer, and junk when the check finds no business. It reads rejected once a removal request has been acted on, which is what takes it out of the feed and out of every export. | BasisOur own run |
| Whether it is a business | Where it comes fromAn AI model reads a website's address, title, description, language, menu, the kinds of structured data it declares, a few counts such as how many prices and forms it shows, and its first few lines, with any email address or phone number removed first. It says whether the site belongs to a business or to a person selling their work. Blogs, news sites, churches, charities, clubs and parked pages stay out of the feed. Websites found before this check existed are checked too. We keep the answer, the model's name, the date it answered and the date it last tried. | BasisRead from the page |
| Two flags and a score | Where it comes fromWhether an address was found, whether any social profile was found, and a number out of 100 for how much of a real business the page looks like. All three are worked out from the rows above, and none of them adds anything new about the business. | BasisWorked out from the above |
What we never do
Getting removed
One email, no account, no reason required. The removal page writes the message and your own mail app sends it, because the address it comes from is the whole check. It applies at ingestion, at unlock and at export, and it holds through every future run.