Regulators increasingly publish their registers as downloadable data. Turning those files into an accurate picture of who operates what is harder than it looks, and the lessons reach well beyond the industry where we learned them.
Open data comes with a reassuring promise: publish the records, and anyone can check the facts. Company registers, financial-services registers and licensing databases are now routinely available as downloads or APIs, and a growing number of products are built on top of them, from know-your-business checks in fintech to supplier due diligence and brand-safety tools.
What the promise leaves out is that a register row is not yet a fact about the world. It is a claim, filed for a legal purpose, in a shape that suits the regulator. We learned this in detail while mapping every online casino brand licensed in Great Britain to the company that actually operates it, and to its sister sites: the other brands running under the same licence. The industry is unusually well documented, which makes it a good laboratory: if the data is messy here, it is messy everywhere.
The dataset
Great Britain’s Gambling Commission publishes its licence register as downloadable files. The one that matters most for the web is a plain three-column table: licence account number, domain name and status. The status takes one of three values: Active, White Label or Inactive. A snapshot we took in September 2026 held 2,241 rows, of which 1,064 were Active, 376 White Label and 801 Inactive. A companion file lists the trading names registered to each account in the same format.
It looks as if “who runs which website” is a single join away. It is a starting point, nothing more.
Lesson 1: a domain is not a brand
The first instinct is to count rows. That produces numbers that are precise, easy to chart and wrong.
Take evoke plc, the group behind 888 and William Hill. Across its four licence accounts the register lists 92 domains, 17 of them live. A customer, however, can open an account at five casinos: 888casino, 888poker, 888sport, 777 and William Hill. The rest are portals such as 888.com, 888.co.uk and 888.info, extra domains for the same poker and sports products, and marketing microsites.
Platform businesses behave differently. Skill On Net Limited, which operates PlayOJO, has 58 live domains that resolve to 50 brands, because seven casinos, PlayOJO and Spin Genie among them, hold two domains each. Anyone listing the Spin Genie sister sites from that licence should count 50 brands, not 58 websites. Jumpman Gaming Limited, which builds and runs sites for partner brands, sits at the other extreme: 149 live domains and 148 casino brands on a single licence, where the only duplicate is one casino registered under both a .co and a .co.uk address.
The ratio of domains to brands turns out to be a property of the business model, not a constant you can correct for. The engineering consequence is simple to state and tedious to do: normalise the obvious noise first (protocols, “www.” prefixes, trailing slashes), then resolve entities with a human in the loop, and never show users a count of rows as if it were a count of businesses.
Lesson 2: the licence holder is not always the name on the door
The White Label status exists because many brands do not hold a licence of their own. They trade on another company’s licence, platform and compliance team.
That produces relationships a flat “owner” field cannot represent. The casino trading as Genting Casino is registered to Skill On Net Limited’s account, and so is Slingo. Genting’s own UK licence account holds no web addresses at all; its online casino lives on the platform operator’s licence instead. In the same way, the licence behind William Hill’s betting shops lists no domains, because the online business sits on separate accounts.
For anyone building on registry data, the lesson is to model a chain rather than a pair: the brand a customer sees, the platform that runs it, and the licence holder that answers to the regulator. Fintech teams will recognise the pattern from know-your-business work, where the entity signing a contract, the one processing payments and the one holding the licence are frequently three different companies.
Lesson 3: a register describes the present, so keep the past yourself
The Inactive status is quietly the most informative column in the file. It is where brands go when they close, rebrand or leave a platform.
Jumpman Gaming’s account carries 80 inactive domains alongside its 149 live ones; Skill On Net’s carries two. Neither number says how well either company treats its customers, but together they describe very different estates: one where partner brands arrive and depart constantly, and one where brands tend to stay put.
Registers rarely publish their own history, so the only way to see change is to keep dated snapshots and compare them. A new domain appearing under an account is an early signal of a launch; a block of domains flipping to Inactive can mark a rebrand or a market exit long before any announcement.
Lesson 4: pair the record with evidence from the real world
A licence register tells you who is authorised. It does not tell you what a website currently offers, what its terms say or whether its claims are true. In a regulated market those details change on a schedule set by the regulator. When the Gambling Commission capped bonus wagering requirements at 10x in January 2026, operators across the market rewrote their promotional terms, and any dataset describing offers from before that date quietly went out of date.
The robust approach is to treat every statement as a claim with a source and a date, and to grade it by the kind of evidence behind it: something observed directly on the live site, something stated by the operator, or something reported by a third party. That way, readers can see how much weight any single statement can bear.
What this means beyond gambling
None of these problems is unique to casinos. Any team that turns public registers into product features runs into the same traps:
- Fintech and payments: onboarding checks that match a trading name to a licensed entity need the brand, platform and licence-holder chain, not a single owner field.
- Marketplaces and procurement: supplier due diligence that counts registrations overstates how many independent vendors really exist.
- Brand safety and security: lookalike and redirect domains are exactly the noise that entity resolution has to separate from real brands.
- Research and journalism: market-concentration figures built on row counts can be wrong by a factor of three or more.
A short checklist for working with register data
- Store every raw download unchanged, with the date you fetched it.
- Normalise mechanically: letter case, protocol, “www.” prefixes and trailing characters.
- Resolve entities with human review, and record why any two rows were merged.
- Model relationships as a chain of brand, platform and licence holder, not a single owner.
- Compare snapshots over time to capture the launches, rebrands and exits the register will not tell you about.
- Attach a source, a date and an evidence grade to every public claim.
- Never publish a count you cannot defend row by row.
The register is the beginning of the answer
Open registers are one of the most useful things regulators do, and nearly every figure in this piece comes from one. But the value is not in the download. It is in the careful, sometimes manual work of turning filings into facts, and in being honest about which of those facts were checked and how.
