Rating method

1 · Principles

Guidance without advice is a construction, not a disclaimer

Four of the seven editorial principles decide how ratings work. A feature, a text or an assistant answer that breaks one goes back before release.

1. 1

   Properties, not persons

   We rate products and describe assets. We never rate people, portfolios or life situations, and never ask about age, income, wealth, horizon or risk tolerance.

   “How important is control to you?”, never “How old are you?”

2. 2

   Public, not personal

   Every result is reachable by anyone through a URL and computed from the published method and visible weights. No account individualises a result.

3. 3

   Teach, don’t select

   We provide the vocabulary, the mechanics, the data and the comparison tool. The reader makes the selection.

4. 4

   Show all sides

   Every product has strengths and weaknesses, shown with the same weight. A ranking says how products rank under a weighting, nothing more.

2 · Criteria catalogues

Criteria and default weights

The default weighting is a published convention, not a judgement of what matters to a particular reader. On a comparison you set your own.

Frontier: the highest score for the criterion among the rated, current products of the category in this release.

3 · Whole numbers

How evidence becomes a whole number from 0 to 100

overall = round_half_even( Σ(wᵢ · sᵢ) / Σwᵢ )

4 · Frontier and confidence

Good in absolute terms, or the leader of a weak field?

Every value is shown against the frontier: the highest score for the criterion among the rated, current products of its category. 82 / 96 leaves room; 82 / 85 is close to the highest measured.

Confidence is a pattern and a word, never a colour. The lowest confidence label among the criteria with a weight above 0 and a score. It changes with the weighting: a criterion with weight 0 does not lower it.

5 · Versions

Every change to the method is versioned and explained

A new version recomputes every category. Every rating names the method version it was produced under, and published versions are never edited.

1. Version 1.2.0Newest release 15 Oct 2026Current

   1 release

   What changed in version 1.2.0

6 · Rating committee

Qualitative points are decided by a committee, on the record

The rating committee reviews scores, reassessments and disputes. Its minutes and decisions are published in the public reference repository and reach this site with each signed release.

7 · Disputes

Anyone can challenge a rating, with evidence

Makers and readers use the same form, without an account. Each dispute gets a public reference and a status page. Nobody sees a draft of a new rating, including the person who raised it.

1. Stage 1

   Received

   The dispute gets a public reference, with the product and the criterion in question. Evidence files are checked for their type and stored privately.

2. Stage 2

   Under review

   Support accepts it for review, and the rating committee reviews the statement and the evidence. The value stays published while the review runs.

3. Stage 3

   Decided

   The decision and its reasoning are published with a reference release. Any change goes into the changelog as a dispute decision.

Challenge a ratingAll disputes →

8 · Wording guide

The advice boundary is drawn by words

The wording guide lists permitted and prohibited terms per language. It is enforced as a lint rule on every interface text and checked on every assistant answer and every published page.

Every feature also passes a four-question test before release: does it select something for the reader; does it concern a transaction or a service; is it presented as suitable or based on personal circumstances; is it addressed to a specific person? One yes sends it back.

9 · Bias review

A rating system in which Bitcoin can never lose points is not a rating system

The founding team’s affinity for self-custody is a motivation and a risk. Before each method version and each major content release, a reviewer with the opposite standpoint, from banking, a central bank or consumer protection, checks where criteria or explainers favour one side, and where bank money and custodial products have not been given their due where they are strong.

How we know

The method is an open repository

Methodology, catalogues, rating rationales, committee minutes and the scoring code are public, so anyone can recompute a ranking and propose a correction.