The UK Sponsor Register CSV: What It Contains, and What It Doesn't

The register of licensed sponsors is free. Anyone can download it, and if your question is "does this one company hold a licence", the free file answers it in about thirty seconds.

Most business questions are not that question. This article sets out precisely what the published file gives you, and what has to be built on top before it supports analysis, targeting or monitoring.

What is actually in the file

The published register is a CSV with a small number of columns per row:

Column	What it holds
Organisation Name	The sponsor's name as registered
Town/City	A single town or city
County	County, often blank
Type & Rating	The tier and rating, for example Worker (A rating)
Route	The route the licence covers, for example Skilled Worker

That is close to the whole of it. It is a flat list of who holds what licence, published as a snapshot.

The rating is worth understanding because it is one of the few genuinely evaluative fields present. Sponsors are shown as A rated, B rated or Provisional. A B rating means the sponsor is on an action plan following a compliance concern and cannot assign new certificates until it is resolved.

The five gaps

1. One row per licence, not per company

An organisation licensed for more than one route appears more than once. Counting rows gives you licences, not companies, and the two numbers differ substantially. VisaAtlas currently tracks 127,283+ unique companies across 143,159+ sponsor records, which is the size of the gap on a single snapshot.

Any headline figure built by counting rows in the raw file will overstate the number of sponsoring organisations.

2. No industry classification

The current published format has no industry or sector column. If you want to know how many care providers, NHS trusts or software companies hold licences, that classification does not exist in the source and has to be derived.

Deriving it is not a lookup. Organisation names are inconsistent, many are holding companies or trading names, and the route alone is a weak proxy for sector.

3. Location is a single town string

There is a town and sometimes a county. There is no postcode, no region, no full address, and no normalisation. "London", "LONDON", "Greater London" and a London borough name are different strings describing overlapping places.

For a company with sites in several cities, the register generally shows the registered or head office location rather than where the work happens.

4. No sponsorship volume

This is the largest gap, and the one most often missed.

The register records that a licence exists. It records nothing about whether it has ever been used. A company that sponsored two hundred workers last year and a company that has held a dormant licence for three years without assigning a single certificate look identical in the file.

For a job seeker that difference determines whether an application is worth making. For a recruiter or an analyst it determines whether the organisation is a real target at all. We cover it separately in how to tell an active sponsor from a dormant licence.

5. No history

Each publication is a snapshot that replaces the last one. There is no version field, no added date, no removed date, and no changelog.

If you want to know which organisations were added this month, which disappeared, or which moved from A rating to B, you have to have kept the previous file and diffed it yourself. Nobody who starts using the register for the first time has that history, and it cannot be reconstructed after the fact.

What "cleaning it" actually involves

Teams usually underestimate this, because the file opens fine in a spreadsheet and looks tidy.

The work is:

• Deduplication across licence rows to a company entity, which requires name normalisation because the same organisation is not always written identically.
• Name normalisation, handling legal suffixes, ampersands, punctuation and casing, so that "A B C Care Ltd", "ABC Care Limited" and "ABC CARE LTD." resolve together.
• Location normalisation to a consistent city and region.
• Industry derivation, which needs an external signal, since the source has none.
• Version storage, so that today's file can be compared with previous ones.
• A join to sponsorship volume data, which lives in an entirely separate publication with its own naming conventions.

None of that is exotic, and a competent engineer can do all of it. It is simply several days of work before the first question gets answered, repeated every time the source is republished, on a file of well over a hundred thousand rows.

When the free file is the right answer

Be honest about this. If you need to check a handful of named employers, or you want a one-off count for a blog post, download the free CSV. It is authoritative, it is current, and paying for a cleaned version of it would be silly.

The free file stops being sufficient when you need any of the following:

• counts of companies rather than licence rows;
• filtering or segmentation by industry or region;
• a distinction between active and dormant sponsors;
• change tracking between register versions; or
• the same view maintained repeatedly, rather than once.

Where our dataset fits

VisaAtlas Data is the cleaned, deduplicated and enriched version of exactly this file, joined to Home Office sponsorship transparency data so each employer carries real certificate-of-sponsorship context rather than licence status alone. Each version records its generation date, register source date, record counts, coverage and a SHA-256 checksum.

It is not a replacement for the official source, and it is not privileged access. Everything in it derives from public data you could assemble yourself. What you are buying is the assembly, maintained.

See what is included in VisaAtlas Data →

Sources and verification

This article was checked on 26 August 2026 against:

• GOV.UK: register of licensed sponsors, workers;
• GOV.UK: sponsorship transparency data; and
• Sponsor guidance Part 1: apply for a licence, for ratings and licence types.

Sponsor counts shown here come from our processed copy of the published register and change as new versions are released. The published format itself has changed before and may change again, so confirm the current columns against the official download.