How we score 67 banks. Out loud.
Every bank in our ranking gets a single number out of 100 — the weighted average of six scored pillars. No editorial overrides, no affiliate weighting, no "we couldn't decide so we made it up." Here's exactly how each cell is calculated, what we won't score, and how we update.
The six pillars
Sum: 100% · weights frozen each editionWhat we actually measure.
Every pillar breaks down into named sub-cells with a defined point ceiling. The score is the sum. There is no "editorial bonus." If you tell us which cell is wrong, we can argue about that single cell — not vibes.
The two largest pillars — Regulation and Fees — together account for half of the score, because in our reader research these are the two dimensions readers most commonly get burned on. Below is the full Regulation pillar at the cell level.
| Sub-cell | Source | Pts |
|---|---|---|
| Licence class | regulatory_tier + licence_strength_score | 12 |
| Regulator quality | regulatory.primary_regulator | 8 |
| Sanctions history (last 36 months) | editorial / news collection | 6 |
| Capital ratios (when published) | annual report disclosures | 4 |
| Pillar total | Regulation | 30 |
Each pillar has a similar table — readers can audit any score by adding sub-cells and confirming they match the headline number. We publish per-bank scorecards alongside every long-form review.
What we refuse to score.
Some criteria are plausible-sounding but unmeasurable, gameable, or actively distortionary. We've explicitly excluded them. Listing what's out is more honest than pretending the inputs are exhaustive.
Brand sentiment. A Twitter scrape of "people complaining about bank X" is not data. Customer complaints reach us via two channels we trust: app- store ratings (anonymised, aggregated, weighted by recency) and our own quarterly support-ticket tests.
Affiliate-weighted "deals." When we link to a bank we earn a commission — disclosed, ledgered, audited. That commission has zero input into the score. A €0-commission bank can rank #1; a €120-commission bank can rank #40.
Marketing claims. "10 million users" is a press-release input, not a score input. Reach in our methodology means licensed-to-operate jurisdictions, not press releases.
"Innovation." A vague catch-all that mostly means "we like the founder." Out.
A worked example: Revolut, Edition №08.
Below is the per-pillar breakdown for the highest-scoring bank in the current edition. The math is reproducible — every number derives from the rubric tables, not editorial judgement.
Revolut · per-pillar
Edition №08 result
Each pillar's contribution = (subscore × weight) / 100. Sum across all six pillars produces the headline number — same formula applied to every bank in the index, no exceptions.
30 + 15 + 19 + 12 + 10 + 5 = 91.
"We score banks the way we'd want our own deposits scored — by what's documented in the licence file, not what's in the launch deck."
Editorial board · Edition №08When the numbers move.
Scores are updated quarterly at edition cutoff. Between editions only two things change a published score: a regulator action (fine, licence change, growth cap) or a verified product change that touches a measured cell (e.g. monthly fee changes). Both are logged in the news timeline first; the score follows in the next edition.
The methodology itself — the weights and sub-cell definitions — gets reopened once a year, in the Q1 edition. Below is the changelog.
Sources, corrections, conflicts.
Primary sources only for regulator data: BaFin / ECB / FCA / Finanstilsynet / ACPR / SARB / MAS public registers, plus banks' own legal-entity disclosure files. Secondary sources (Reuters, FT, Bloomberg) get cited but never carry a score impact alone.
Corrections. When we get a cell wrong, we publish a dated correction note on the bank's review page and in the next edition's changelog. The point we deducted (or added) in error is restored within 24 hours of verification — even if the next edition is two months out.
Conflicts of interest. Contributors who hold a personal-investment equity stake in a bank we score are recused from that bank's score. The recusal note appears in every edition that bank participates in.
How we get paid (and why it doesn't move the score).
We earn affiliate commission when a reader opens an account through a link on this site. Commission ranges €0 to €120 depending on the bank. Three things keep this from corrupting the ranking:
One. The score is calculated before the affiliate ledger is loaded into the layout. There's a literal git boundary — the editorial branch produces the scorecard; the commerce branch wires up the buttons.
Two. The full disclosure ledger — every commission rate, every quarterly payout, every contributor whose salary it funds — lives at /disclosure/. That's how we publish "Total Q1 affiliate revenue: €18,420 · funds 2.4 of our 6 contributors" on the homepage without flinching.
Three. Banks that don't pay us are scored anyway. A meaningful share of the index has no affiliate relationship at all; some sit in the top 10 regardless.
Spotted a cell you think is wrong? Tell us.
Corrections are read by an actual human editor, usually within a week. We re-score and credit you in the changelog.