How Skrumble scores crypto platforms

It summarises documented evidence and compares it across the products we cover in one category. A wallet scored 4.4 has stronger evidence behind it overall, once the five pillars are weighted, than one scored 3.9, though it can still trail on a single pillar. A score is not a price prediction, a guarantee of safety or advice about what you should buy.

Read the pillars, not just the total, on any wallet or exchange review. Two products with the same overall score can be strong in different places, and which pillar matters most depends on you. Every scored review shows its pillar scores next to the overall, and every category page shows a worked example of the arithmetic.

Scores only compare like with like. An exchange score and a wallet score use different pillars and weights, so an exchange at 4.2 and a wallet at 4.2 say nothing about each other.

Evidence, for Skrumble, is anything we can link to and date. We rank it in this order, and a higher source beats a lower one when they disagree:

1. Regulators, registers, statutes and court records, such as ESMA's interim MiCA register, the FCA register, US state licence pages and court dockets.
2. The company's own documents: terms of service, fee schedules, pricing disclosures, cardholder agreements, help-centre pages, audit reports and public code.
3. Reputable press, such as Reuters, Bloomberg, CoinDesk or The Block.
4. Data aggregators such as DefiLlama and CoinGecko, for market data (supply, volume, value locked) and incident tallies, always named as the source.

In reviews, other review sites, affiliate blogs and comparison tables never count, even when they happen to be right. Marketing numbers such as user counts are reported as the company's own claims. Every source in a research file records the day we opened it and, when the page shows one, its own date, so stale pages are visible as stale.

Every fact is checked through a research file that never appears on the site but drives everything on it. Since October 2026 each review's file is built in four passes, and a page that fails any of them does not ship:

1. Claim ledger. Every factual statement (fees, dates, licences, incidents, user numbers) goes into a table with its current value, the source URL, the publisher, the source's own date and the date we opened it. We cut any claim that has no fetched source.
2. Adversarial challenge. A separate second pass re-opens the sources for every high-stakes item (money, licences, incidents, lawsuits) plus a random 30% of the rest. Where the two passes disagree, the challenge wins.
3. Sentence-level QA. On the finished draft, every sentence that states a fact must map to a ledger row or a fetched source. We fix or delete any sentence that does not map.
4. 100-point audit. The audit scores the finished page out of 100 on content quality, search optimisation, sourcing and expertise signals, technical structure and how easily search and AI tools can cite it. Below 75, the page does not ship. The target is 85.

Each pass leaves its own record: the claim ledger, the challenge notes, the sentence map and the audit report. We keep them on file for every page that went through this process, so any of its sentences can be traced to the document behind it. Each category page shows a real example of something the challenge pass caught.

How did the process change in October 2026?

Pages published under the May 2026 version were drafted first, and then each page's highest-stakes claims, usually 10 to 20, were verified against primary sources. On the Binance, BTCC, Bybit and Kraken reviews that check covered the scores and data fields; their long-form text was carried over from the old site and has not been re-verified. The other exchange reviews were rewritten in May 2026. Nine exchange reviews described tests we have no record of, such as order timings and support response times; those passages were removed on October 17, 2026. Since October 2026 every claim goes into the ledger before writing starts, a separate pass challenges it, and every sentence is mapped after writing. 27 of the 52 published reviews have been through the full October process. 9 bridge reviews were refreshed in October from the claim ledger and challenge pass behind our bridges guide, without a sentence check and audit of their own. The rest still show their earlier date. Each category page, and each stablecoin review's date, shows which version applies.

No. Skrumble does not open accounts, buy devices, deposit money, place trades, apply for cards or contact support teams as a customer. Every review is documented desk research. Where a review reports a quote or a route read from a public interface, such as a bridge's API, it was read without moving funds and it is dated.

Desk research has limits, and you should know them. A review cannot tell you how long your own identity check will take, whether your bank will decline a payment, or how a support agent will treat you. It can tell you what a company has committed to in writing, what regulators and courts have recorded about it, and how that compares with its competitors.

Correction, 17 October 2026: earlier versions of this page and of the category method pages described hands-on test procedures, such as test purchases, funded accounts and support tickets, and showed a "Tested" label. We have no record that those tests were carried out. We removed the descriptions and the label and rewrote the method pages to describe the work as it is actually done. The same kind of wording also appeared in older exchange reviews, a guide, review titles and labels elsewhere on the site, with no record of the tests behind it. We removed it on October 17, 2026; where a sentence depended on a test, we deleted it rather than replace it with a claim we could not source.

Reviews, most country guides and the category guides on each hub are published under Matilde Ferreira's byline. Most coin explainers carry Carla Moretti's. Method pages and many explainers are published as Skrumble Editorial. Each author page lists the reviews and guides published under that name and the topics it covers.

Google's guidance asks publishers to think about who created a page, how and why. The author pages say whose byline a page carries, and this page sets out the method.

Every review shows the date it was last updated. Our target is a full re-check of every published review at least every six months, and sooner when something material changes:

• a fee, price or rewards change
• a licence or registration granted, refused or withdrawn
• a hack, an exploit or another security incident
• a product closing, or leaving or entering a country
• a court ruling, a fine or another regulator decision

Each category page lists the triggers that matter most for that category. A review older than six months is not necessarily wrong, but check its updated date, and the dates in its source list where it has one, before you act on a number.

Tell us through the contact page: the page, the claim and the source you think contradicts it. We re-open the primary source. If we were wrong, we fix the sentence, re-check any score the fact feeds, move the page's updated date and record the change, with its source, in the research file.

A corrected fact does not always move a score, because each pillar is judged against the other products in the category. When a score does change, the tables on these method pages change with it, because they read the same data as the reviews.

Skrumble's business model is affiliate commission. A platform can pay us when a reader signs up or buys through a partner link. Today no outbound link on a review carries a referral code. The Visit buttons go to the companies' own websites. The Visit buttons on reviews are marked rel="sponsored", one of the values Google lists for qualifying paid and affiliate links.

Money cannot change three things:

• Scores. The overall score is computed from the pillar scores and weights alone; commission terms are not an input anywhere in that calculation.
• Inclusion and order. Country guides list only products with a published review, and their order is an editorial judgement explained in each entry, not a ranking by commission.
• Content. No company approves a review before it is published, and none can pay for placement.

Disclosure rules point the same way. The US FTC tells publishers to "disclose your relationship to the retailer clearly and conspicuously", and the UK's CAP Code requires marketing to be "obviously identifiable as such". Our full disclosure is on how we make money.