Crypto.com

Its proof of reserves uses the right method and the wrong date. The Merkle tree is genuinely verifiable — you can check your own balance is included, which is more than several larger venues offer — and the page presenting it names an attestor that stopped doing cryptocurrency work in December 2022, still citing that date as the attestation.

That is not a technical quibble. The value of an attestation is that an independent firm looked recently. A proof whose named reviewer has not been in the business for years is a proof of what was true then, presented as though it were current, and the fact that the page has not been updated tells you how much attention the disclosure receives internally.

The insurance figure is misread even more often. The widely cited "$750m" is a 2021 cold-storage policy — cover for assets held in cold storage under defined conditions, not a fund that pays a customer whose balance disappears. Almost every article quoting the number treats it as the latter.

The weakest part of the rating is KYC and limits at 3 out of 10: the published thresholds and tier structure are unclear in ways that matter if you are moving meaningful sums and want to know in advance what will be asked of you.

None of this describes a venue that has lost customer money. It describes disclosure that has been allowed to go stale, which is a different and more ordinary problem — and an easily fixed one, which is why it is worth naming.

A $10,000 bitcoin buy cost 0.25% on 14 July 2026 — two and a half times what Binance, Bybit and Bitget charged on the same day, and just under Kraken’s 0.26%. Unlike Kraken, there is no proof of reserves attested by a current firm to justify the premium.

Depth is thin. We measured 30.1 BTC within 0.1% of best ask, a sixth of Kraken’s book and half Bybit’s, so larger orders move the price further here. Effective cost at $100,000 was 0.2523%.

Infrastructure measured worst in the run: 572 ms of API response latency, more than three times Coinbase’s 183 ms. For an app-first retail venue that matters less than it would for a trading venue, and it is still last.

What people are usually buying is the app, the card and the onboarding, which are genuinely good and are not what we measure. On the numbers we do measure — what a trade costs, how deep the book is, how fast the API answers — it is mid-table at best, and paying a premium for nothing the competition lacks.