How we rate and report

Every score we publish is computed, not typed. The goal is that you can recompute it yourself from the evidence on the page, and that two analysts scoring the same provider independently land within 0.3 points. This is rubric version 2.0.

The five pillars

Each category is scored on the same five pillars, so scores stay comparable without pretending a hardware wallet and a tax tool are the same kind of product. Only the indicators and the weights change.

Counterparty and custody risk — if this provider fails, what happens to your assets?

Cost transparency — what it charges, and how much of that you can read before opening an account.

Market quality — depth, liquidity and longevity, taken from an independent public dataset rather than our own measurements.

Transparency and track record — what is disclosed, and what has happened before.

Public documentation surface — what a prospective user can read before signing up: fees, status, API, legal terms and help.

Weights follow one principle: the more of your money someone else can lose, the more of the score is risk. Bridges and stablecoins carry 45% risk weighting; crypto cards, which are a payment rail, carry 25% risk and 35% cost.

How a pillar becomes a number

Each pillar decomposes into four to six indicators with stated thresholds. An analyst records whether each one is met, partially met, or failed — scoring 2, 1 or 0 — and the pillar score is points earned divided by points available, on a ten-point scale. The overall score is the weighted sum. Nobody enters a score directly.

Every indicator carries a source and the date it was checked. Where a primary source exists — a regulator's register, an auditor's report, a public dataset, the provider's own legal documents — a marketing page does not count as one. Where a provider publishes a page at a documented URL but blocks automated retrieval, the indicator scores partial with the block recorded, never a failure: we grade exchanges, not firewall configuration.

What we deliberately do not score

Country availability is out of scope. We do not score whether a provider serves your jurisdiction, and gaining or losing access to a particular market never moves a score. What we do score is whether the provider holds authorisation from a recognised financial regulator and publishes it, which is the same fact wherever you happen to live. Where availability is materially restricted we say so in the review — a reader needs to know it, but it is information rather than a deduction.

We also do not score anything we cannot verify ourselves from a public source. If an indicator would require an account, a private disclosure or our own market measurements, it is not in the rubric at all rather than scored as a failure.

Gates

Some failures are not tradeable against convenience, so they cap the score outright rather than being averaged away. A provider with no identifiable operating entity cannot exceed 5.0. A custodial platform with no reserve attestation in the last twelve months cannot exceed 7.0. An unremediated security incident with unreimbursed user losses caps at 6.0, an undisclosed withdrawal suspension at 6.5, and contracts holding user funds behind a single upgrade key with no timelock at 6.5. Sanctioned entities, and anyone who asks to pay for placement, are excluded and we say so on the page.

Freshness and confidence

A score is a claim about a date, so we publish the date. Inside 90 days a review has normal standing. Between 91 and 180 days it is flagged for re-check and becomes ineligible for Editor's Choice. Between 181 and 365 days it sorts below every verified peer. Past a year it is unpublished until re-verified.

Alongside the score sits a confidence grade — A, B or C — set by the share of indicators verified from primary sources. A grade C score is published with the grade visible and is never presented as a recommendation. Editor's Choice requires grade A and a verification date inside 90 days; a high score on its own does not qualify.

Independence

Scores are produced before any commercial conversation. A partnership never changes a score and a refusal to partner never changes one either. Affiliate links, where present, are disclosed and have no input to ranking order. Any score movement of 0.3 or more requires a dated entry naming the indicators that changed and the evidence behind them, published on the review. An analyst with a position in, or a relationship with, a provider does not score it.

Rubric changes are versioned, and every score records the version that produced it. When a rubric changes, the affected category is re-scored before the new version goes live — scores from two different versions are never shown in the same table. Providers may dispute an indicator by supplying evidence; we re-verify, correct where we were wrong, and record it. Disputes never pause a published score.

Market analysis

Editorial analysis is separate from ratings. It draws on primary sources first — the provider's own disclosures, on-chain records and live market feeds — cross-checked against independent references, and we publish the reasoning rather than only a conclusion.