UKv2.1
The principles behind every IndexFair rating — the same across every market we cover.
IndexFair is an independent analytical layer over public consumer reviews and first-party product measurement. We don't host reviews, we don't take payment to change a score, and we publish the framework. Each market has its own detailed methodology; this page is what they all share.
Every IndexFair number is a country-and-category-relative index, from 0 to 10, attached to one brand, in one market, in one country. It ranks the public evidence we could independently verify for that brand against its peers in the same country and category — it is not an objective measure of the operator's true quality, and not a guarantee of any user's future experience. It is the deterministic output of a versioned formula: given the same inputs and the same methodology version, the same number reproduces.
Each index carries a methodology evidence range and a percentile against the brand's market. A wider range means a smaller or more divided evidence base, so the index is less precise — not that the operator is worse. Where the evidence is too thin to be fair, we withhold the number rather than publish one we cannot stand behind.
What we check. Identical inputs under the same methodology version must reproduce the same output. Source-bound facts must clear factual-accuracy and conflict checks. Sensitivity, robustness, cohort-stability, and dependency-group tests must pass, and the exact configuration must complete at least two flagged dark refreshes across 14 calendar days.
What we do not claim. We do not claim that the ordering has been validated against real-world outcomes or an expert panel. Reviewer-agreement studies and outcome backtests are optional research commitments, not conditions for publication. If we run them, we publish the protocol and result.
Six principles shape every score, in every market:
We read real user reviews across app stores, social and community platforms, review sites, and specialist pages — never the marketing copy. A published score draws on at least two distinct data sources; the full source mix is not disclosed, so the input cannot be reverse-engineered and systematically targeted.
Every review passes a deterministic filter — duplicate fingerprints, posting bursts, length and sentiment anomalies — before it can affect a score. Reviews that fail are excluded, and their count is shown openly on each brand page.
We independently estimate each brand’s real-world traffic and how its users arrive, as a corroborating signal: review sentiment that is not backed by a matching audience is trusted less.
Where a product can be measured directly, we measure it ourselves rather than take the operator’s word — and feed those facts into the score as their own block.
Evidence decays with age: a report from last month counts for more than one from three years ago, on a published half-life curve.
Small or divided samples are pulled toward the market average and carry a wider methodology evidence range; below a minimum threshold the number is withheld rather than shown at low confidence. The range describes how much evidence we have, not how good the operator is.
The gambling composite combines four independent blocks. Each contributes a weighted sub-score; the blend is the published number.
A 0–10 score is an evidence index within one scale — not a universal quality measure. Scores are comparable only within the same category, jurisdiction and methodology major version. A sportsbook 8.0 and a casino 8.0 are different constructs — margin transparency versus game breadth versus reserve cover — and are never ranked against each other.
Change any segment and the numbers stop being comparable. The compare tool only opens for brands that share a scale id — there is no global leaderboard and no cross-vertical sort.
Each market keeps these principles but adds its own scoring blocks, aspects, and weights. The detailed, versioned methodology for each lives on its own page:
Bookmakers and online casinos — the four-block composite (trust & compliance, user experience, operational signals, measured product), the per-aspect weights, and the full source and fake-review detail.
Read the gambling methodology →Crypto exchanges — the three-block composite (trust & compliance via FCA registration, user experience, operational signals), the nine exchange aspects, and the regulatory regime. Wallet and DeFi leaves are upcoming.
Read the crypto methodology →VPN services — verifiable facts (jurisdiction, ownership, published audit history) alongside user-reported service quality. No overall safety rating is published in Phase A.
Read the vpn methodology →When a brand we rated highly later fails, our response is governed by a protocol we publish in advance — When we are wrong.
A number from 0 to 10 for one brand in one market, the deterministic output of a versioned formula. Given the same inputs and the same methodology version, the same number reproduces.
We do not accept payment to alter scores, surface review text verbatim, promote brands by ranking position, use marketing language, or recommend a brand by name in editorial copy.
Every review passes a deterministic filter — duplicate fingerprints, posting bursts, length and sentiment anomalies — before it can affect a score. Excluded counts are shown openly on each brand page.
Gambling (bookmakers and online casinos) is live with the full four-block methodology. Crypto exchange and VPN coverage are live as well, each with its own methodology page.
To cite IndexFair, attribute scores and rankings to IndexFair's independent methodology v2.1 and link to the methodology page at indexfair.com/methodology.