What is deliberately not published
This site publishes organisations. It does not publish people. That distinction is not obvious in the source data, which is why it gets its own page.
Municipal licence files do not separate companies from people. Vancouver’s business licence extract returns sole proprietors under their own legal names — the second record the API hands back is a named individual trading as a cabling company. The federal grants file pays individuals alongside universities.
Somebody who takes out a trade licence should not thereby acquire a permanent, indexed, searchable public profile. So every subject is classified before anything about it is published.
The obvious approach — a list of business words — fails, and fails in the direction that matters least to a developer and most to a business owner. A token-matching classifier on a sibling project suppressed 480 of 3,000 names, including a major homebuilder, an established restoration contractor and a design studio. Business names are novel by design: no list of words will ever contain the next company’s name.
| Signal | What it settles |
|---|---|
| Legal form | A name carrying a registered suffix — Inc., Ltd., Ltée, ULC, LLP — is an organisation. This is not a heuristic; it is what the suffix means. |
| Structure | Personal names have a shape: two or three plain words, no digits, sometimes a parenthesised short form of the same name, sometimes an inverted “Surname, Given” form. Company names rarely do. |
| Corpus frequency | A name appearing ten or more times across the corpus is a business. One person holds one licence; a company holds forty. This is the single tunable privacy parameter, and it is exported, tested and documented rather than buried in a helper. |
| The publisher saying so | Outranks all three. The federal grants file carries a recipient_type code that states outright when a recipient is an individual, and Calgary flags home-based businesses directly. An assertion always beats an inference. |
In the current build, 109,356 subjects were withheld out of 1,913,424 resolved — 5.7%. Of those, 33,587 were identified as individuals and 75,769 could not be confirmed to be organisations.
A registered office address is very often somebody’s kitchen table. So address precision is decided per record rather than shown by default:
- Locality only where the publisher flags a home-based business. Calgary is the one source in the registry that states this outright, and where it does, it is believed rather than inferred.
- Locality only where a single publisher asserts the organisation — the case where a registered office is most likely a residence.
- Full street address where two or more independent publishers corroborate the organisation, which is strong evidence of real business premises.
At locality precision the postal code is reduced to its first three characters. A forward sortation area describes a neighbourhood; a full postal code describes roughly one block face, and in rural Canada sometimes one household.
Phone numbers, email addresses, contact names and applicant addresses are dropped between Silver and Gold — they are not in the published records, so no template can reveal them.
Human-written prose is dropped for the same reason with less certainty about its contents. Contract descriptions and licence conditions are free text, and free text has repeatedly been found to carry given names, phone numbers and email addresses in datasets whose schema promised none. There is no reliable way to sanitise it, so it is held for internal analysis and never published.
Charity boards are the one place this site names private individuals, and it does so on purpose, not by accident. The Canada Revenue Agency publishes the directors and officers every registered charity files with it, under the Open Government Licence. It sat disabled in this project for a long time on the reasoning that a searchable national index of who sits on which board is a different product from an entity registry, and turning it on should be a deliberate decision, not a default.
It is on now, on the narrowest possible slate. The source file contains only a name, a position, tenure start and end dates, and the arm’s-length flag CRA itself requires disclosed — no address, phone, or email exists anywhere in it to withhold. A director’s tenure is shown on their charity’s page and nowhere else: it is never indexed by search, never given its own page, and can never become the name of an entity. A director who happens to share a name with a federal vendor cannot pull that vendor’s contract onto their charity, and cannot pull the charity’s record under an unrelated name either — both are structural properties of how records are resolved, not a filter applied afterwards.
This is also why the site refuses to do the equivalent thing with scraped LinkedIn profiles: that data carries no open licence, no consent to redistribution, and no limit on what it exposes about a person, and scraping it is a practice LinkedIn actively and successfully sues over. The bar here is the same one every other source clears — a government publisher, an explicit open licence, the narrowest slate the source actually contains — and LinkedIn data clears none of it.
Canada’s own federal beneficial-ownership register makes the same choice for the same reason: it can be searched by a company’s name, and its designers deliberately blocked searching it by a person’s name. Every federal corporation’s page here links directly to that register instead of scraping or redistributing it — the link sits next to that corporation’s identifiers.
Every record links to the publisher it came from. Corrections to the underlying facts have to be made there — this site holds no authority to change a registrar’s record, and a correction made only here would put the two out of step. Where a record here is assembled wrongly — two companies merged, or an award attached to the wrong organisation — that is a defect on this side and worth reporting.