How the data is built

How the numbers on this site are produced, in enough detail to check them, including the parts that are guesses and the parts that can be wrong.

Where the data comes from

Current licence records come from the Home Office Register of licensed sponsors (workers), published on GOV.UK under the Open Government Licence. Company details come from the Companies House monthly data snapshot, also open data.

Nothing is scraped from job boards, bought from a broker, or entered by hand. There are currently 127,826 employers on the register.

How often it updates

The register is re-downloaded once a day. GOV.UK does not publish at weekends or on bank holidays, so a day with no changes is normal and not a failure. The Companies House snapshot is monthly, so company details can be up to five weeks behind.

Every page shows the timestamp of the data it is built from. If that date looks stale, it is: something has broken and you should check GOV.UK directly.

How employers are identified

An employer is identified by name plus town, not name alone. Chains and franchises repeat one name across many towns, and treating those as a single employer would hide real additions and removals.

The consequence is worth knowing: if an employer moves town, or the register changes how its town is spelled, it can look like one employer left and a different one arrived. That is a limitation of the source, not something we can fix from outside it.

Industry tags are derived, not Home Office data

The official register contains no industry field. The tags on this site prefer a specific Companies House SIC code where the sponsor name has a safe company match. Ambiguous matches and vague SIC codes are withheld. Where that evidence is unavailable, conservative keyword rules over the sponsor name provide a fallback.

About 12% of employers end up in "Other", because their name carries no clue about what they do. They are there to make searching possible, and they are wrong often enough that you should not rely on them for anything that matters. Employer pages label the classification basis. Always confirm with the employer.

Companies House warnings, and how they can be wrong

Employers are matched to Companies House by normalised name: case, punctuation and legal suffixes are standardised, so "The Foo Company Limited" and "FOO CO LTD" match. Different legal forms are kept apart, because "Foo Ltd" and "Foo LLP" are different companies. Previous company names are matched too, since the register is slow to follow a rename.

This is a name match. It is not an official link, and it is the weakest step in the pipeline. Two things follow from that:

• A warning is not proof. Every warning links to the Companies House record so you can check it yourself rather than take our word for it.
• No warning does not mean vetted. Roughly a third of employers on the register are not limited companies at all. Councils, NHS trusts, schools, universities and charities have no Companies House record to check.

Where one normalised name is held by more than one company and we cannot tell which is the employer, we show no warning at all. Dissolving a company frees its name, so a live employer often shares a name with a dead company, and picking the wrong one would tell somebody their prospective employer was being wound up when it is trading normally. If exactly one of the two is active we use that one, because a dissolved company cannot hold a sponsor licence. Otherwise we say nothing.

Search can also use a current legal name, a Companies House number or a trading name written into the sponsor record. These are discovery aliases, not new identity decisions. Every alias still opens the exact sponsor name and town from the Home Office register, and an ambiguous company match contributes no legal-name or company-number alias.

The warnings are not equally serious, which is why each one says what it found rather than just turning red. A company in liquidation is grave. Accounts filed late is often paperwork at a company trading perfectly well. A company incorporated within the last 365 days is not a problem at all, only unusual enough to mention.

Historical Certificates of Sponsorship

Historical assignment counts come from the Home Office response to FOI 2025/12841. Its annex records Certificates of Sponsorship assigned under organisation names for 2021 to 2024. It covers Skilled Worker and Global Business Mobility / Intra-Company Transfer routes and was extracted from the Home Office operational database on 8 October 2025.

These are assignments, not unique workers, visa grants, current vacancies or promises to sponsor. The response does not split in-country from out-of-country assignments. Values of five or fewer were withheld by the Home Office and appear in the source as an asterisk. We show those values only as 1–5. Where a published total could reveal a withheld value by subtraction, we discard that exact total and publish only a safe range.

The annex has organisation names but no sponsor licence number or company number. We attach a history only when that name is an exact, unique match to the current sponsor register. Similar, normalised and ambiguous names are not attached. Even an exact name is a link between two published names, not proof that the legal entity remained unchanged throughout the period.

2025 CoS use disclosure

A separate Home Office response, FOI2026/03501, reports employer-level counts labelled CoS used for calendar year 2025. It separates Skilled Worker / Tier 2 from Global Business Mobility / Intra-Company Transfer.

Home Office SMS guidance defines assigned and used as different statuses. The sponsor policy says a CoS normally remains assigned while an application awaits a decision. It shows as used after the Home Office grants or refuses the application.

We do not append these figures to the assignment chart. The FOI response does not publish its extraction query or say whether it counted final status or status history. Its annex includes only employer-route counts of five or more. If an employer or route is absent, the result is unknown rather than zero. The figures do not identify an occupation, a worker, an application, a visa grant or a current vacancy. The Home Office also says they came from a live operational database and may change.

The same identity rule applies. A count is attached only when the published employer name occurs exactly once in the current sponsor register. Similar and ambiguous names are withheld.

Damaged characters in the source

The official CSV contains employer names that were already broken before publication: an apostrophe arrives as a byte that cannot be decoded, so a real employer appears with question marks in the middle of its name. The bytes needed to recover them were destroyed upstream, so nobody can restore them, including us.

We repair only the specific damaged shapes we have actually observed, and nothing else. Guessing at the rest would mean renaming an employer to something it is not called, which is worse than showing the damage.

What stops bad data being published

The pipeline runs unattended and publishes straight to this site, so it refuses to publish figures that look implausible. It stops if fewer than 10,000 employers are parsed, or if the register moves by more than 10% in a single run. Real daily movement is a fraction of a percent, so anything near that limit is far more likely to be a broken download than news.

On a refusal nothing is written, the old data stays up, and the job fails loudly so a human looks at it. The site going a day stale is a much smaller problem than the site going confidently wrong.

The change history

GOV.UK publishes who holds a licence today. It never publishes who stopped. So every day's register is compared with the previous day's, and the differences are recorded permanently.

The change log shows the last 30 days, and the homepage highlights the last 7. The full record goes back to when the pipeline started and is never deleted, which is what makes it possible to answer when an employer first appeared, how long it held a licence, and which sectors are growing.

One honest limit: an employer already on the register the first time we looked has no recorded start date. All we can say is "on or before" the day we started watching, and anywhere that date is used it says which of the two it is.

What this cannot tell you

• Whether an employer is hiring. A licence is a permission, not a vacancy.
• Whether they will sponsor you. That is their decision, and it depends on the role and the salary rules.
• Whether a historical CoS assignment led to a visa, whether it belonged to a unique worker, or whether the employer intends to assign another one. The source does not contain those facts.
• Why an employer left the register. Licences lapse, get surrendered, get revoked, and companies close. The register does not say which, so neither do we.
• Anything about your own eligibility. That is regulated advice and this is a search box.

Check our work

The public audit layer contains the source-linked JSON needed to reproduce the core current-register, recent-change and aggregate claims, free and with no account. The full employer-level archive, internal matching rules and customer reports are deliberately not bulk-published. Every displayed conclusion still identifies its source, evidence type or coverage gap. If you think a number is wrong, you can check the documented public files and the linked official record. If you find a mistake, tell me at hello@knowyoursponsor.co.uk.

Last updated 28 October 2026.