About this research
Last updated 4 August 2026
We publish original statistical research on the online gambling industry, built from data we hold because of our commercial work. That last part deserves an explanation rather than a footnote, so this page gives it first.
The conflict, stated plainly
Kasinohai is a casino comparison site. We earn commission when a reader opens an account with some of the operators we cover. That is a real conflict of interest, and it does not go away because the work is careful.
What we can do is make the research checkable, and make the conflict visible enough that you can weigh it. Three things follow from that, and they are the reason this section is built the way it is:
- Studies cover the whole dataset, not our partners. The crypto study covered 2,554 brands; the supply-chain study covered 320,187 relationships. Brands we have no commercial relationship with are in the data on the same terms as brands we do.
- Every chart ships with the table behind it. Each study’s method section sets out its classification rules and exclusions, and the underlying tables are available on request. You do not have to take a number on trust.
- No casino brand or operator is named in any performance context. Not to protect them, but because naming them would turn research into a ranking, and a ranking is a commercial product.
What independence looks like in practice
No operator sees a study before it is published, and no operator has ever been offered the opportunity to respond to a finding before publication. Commercial partnerships are not a factor in what gets studied, and the commercial side of the business has no sign-off over a study’s findings or whether it runs.
A statement like that is easy to write, so here is the harder evidence: findings we have published that work against us.
| Finding | Why it costs us something |
|---|---|
| Six in seven players click through to a casino without opening a single piece of information about it first | The strongest available argument that comparison content, including ours, is not doing the job it claims to do. |
| The advertised wagering multiple does not determine what a bonus costs, and two offers showing the same number can differ twofold | The advertised multiple is a column in our own comparison tables. |
| Only five of 733 modelled bonus offers would meet Finland’s incoming 5× wagering cap | Almost every offer we list would fail the standard the market is about to be held to. |
| German licensing was followed by demand moving further offshore, not onshore | An inconvenient result for the argument that licensing alone protects players, and one we could have left unpublished. |
Where the data comes from
- The structural dataset. A maintained audit of online casino brands: licences, bonus terms, payment providers and game suppliers. Compiled by our editorial team as part of reviewing operators, and covering brands whether or not they are commercially relevant to us. Studies built on it are censuses, not samples.
- The observed-demand dataset. Daily records of how visitors to our comparison sites behave, including which operators they look at, compare and click through to, by market and date, reaching back to January 2017. Powerful but partial: it sees people who shop between operators, not people who go straight to a brand they know.
- Public and regulatory sources. Regulator publications, licensing registers and enforcement decisions, cited in the studies that use them.
The standing rules every study is held to (verification, fixed chart scales, published nulls, and the list of things we will not publish) are set out in the methodology.
Who does the work
Antti Virtanen leads the research and is responsible for the method, verifying every published figure, and deciding whether to publish or drop a finding. He is available to journalists for comment on any study here.
When we get it wrong
We correct published figures and log the correction rather than making it silently. The log also records findings we revised or dropped during verification before publication, including several that would have been wrong in print. You can find it at /research/corrections.
If you think something here is wrong, please tell us. A specific challenge to a specific number is the most useful thing anyone can send, and we check it against the source data rather than defend it.