Methodology · Reviewed Aug 19, 2026 · 9 min read
How We Score Crypto Exchanges: Full Methodology
Every weight, data source and update rule behind our ratings.
Reviewed by CryptoExID Editorial · Aug 19, 2026 · Editorial policy · how we make money
What this page is for
This is the scoring contract behind every CryptoExID ranking. We do not sell placement. A higher score means the model, on the current dataset, prefers that venue for that context. It is not a promise that you will make money, and it is not a guarantee the venue will stay licensed in your country.
We rebuilt the public site around one idea: the same exchange can be the right answer in Germany and the wrong answer in the United States. Availability, rails and the local legal entity change the result. That is why /en/ is the global English model and /en-us/ is the USA model. If two pages share a country, a context and a dataset date, they must print the same score and the same maker/taker pair.
The ten factors and the base weights
Each venue gets ten subscores from 0 to 10. The base (global overall) model mixes them as: Security 20%, Regulation 18%, Liquidity 15%, Fees 13%, Fiat and payments 10%, Features 8%, Asset coverage 6%, UX and mobile 4%, Support 3%, Transparency 3%. Those weights always add up to 100%.
A lowest-fees page does not invent a new database. It shifts weight onto fees and liquidity, then re-filters who is eligible. A safest page shifts onto security, regulation and transparency. A beginners page shifts onto UX, payments, support and regulation. You can read the exact shift on each ranking page, on the line that starts with the model name.
Eligibility comes before the sort
If a venue does not onboard customers in that market, it does not rank there. We still show a short honesty strip of scored-but-excluded names so the page does not pretend the missing brand does not exist. Futures pages drop venues with no futures product. ACH pages drop venues that do not take ACH. No-KYC pages drop venues that require full identity verification before you can trade.
Country pages also rename the local entity when we have one. Binance in the USA model prints as Binance.US. OKX prints as OKX US. The slug stays the same so reviews and visit links do not fork into a second fake company.
Where the numbers come from
Spot maker and taker fees come from the venue's published schedule, checked on the dataset date printed on the page. We use the standard retail tier, not VIP 0.00% marketing. Liquidity is a subscore, not a live order-book widget. Regulation is a subscore based on public registers and enforcement history, not a purchased badge.
We do not currently store per-network withdrawal fees or Lightning flags as first-class fields. If a page needs that level of detail, we say so in prose and send you to the review, where the editorial check date sits next to the claim. Inventing a Lightning tick to make a Bitcoin page look unique would break the contract.
Quality gates, not URL sprawl
A programmatic URL only enters the sitemap when enough venues are eligible and the order is not a clone of its parent. Thin country times intent combinations stay out. That is why the methodology page and the header counter can disagree with an older marketing line that said 1,400 pages. The live count is whatever the registry prints after the gate runs.
If you catch a fee, a licence flag or an availability miss, use the corrections page. We treat that as a data error, not a review argument.
FAQ
Do affiliate links change rankings?
No. Visit links can be affiliate links. The model runs before monetization. Commissions do not move a subscore, a weight or a position.
Why does the USA ranking differ from the global ranking?
The USA page filters venues that onboard US customers and uses local entity names such as Binance.US. The global page at /en/ does not apply that filter. Same database, different eligibility.
How often do scores update?
Whenever we change a fee, a licence flag, a capability or an availability row. The page prints the dataset date. Field-level dates will replace the single build date as the pipeline matures.
Is a 9.0 safer than an 8.0?
Not by itself. A safest-context page already reweights security and regulation. An overall 9.0 can still sit behind a specialist on fees or futures. Read the context line on the page you opened.
Why is a popular exchange missing?
It failed an eligibility rule for that URL: country block, missing product, missing rail, or a no-KYC filter. Check the scored-but-not-ranked strip and the review.
Can I reuse these scores commercially?
You can link to the page. Republishing the table as your own ranking without the methodology and the date is how people get the numbers wrong a week later.