Methodology

This page documents how every ReviewQuorum number is produced, and what those numbers cannot tell you. The current version is 1.1. Every calculation stores the version and the thresholds in force, so old verdicts can always be traced to the rules that produced them.

Four layers, never mixed

• Product facts — manufacturer pages, documentation and certifications. Used for identity and specifications, never for quality claims.
• Buyer Signal — ratings and comments on marketplaces, always shown with platform, region, sample size and date.
• Creator Signal — substantive creator videos and posts, with sponsorship and product-provided flags.
• Quorum Score — qualifying professional reviews only.

The four can appear side by side. They are never added or averaged into a single number.

Which reviews count

A review counts toward the Quorum Score when it is a complete review (not news, first impressions, an embargo hands-on, a spec sheet or a roundup), identifies the product and variant, carries its own editorial judgment, can be traced to its original URL, is not sponsored content, is not a republication, and has been checked by an editor. Reviews that do not qualify can still appear as context, labeled as such.

Independence: the Source Quorum

The Source Quorum counts independent editorial groups, not websites. Publications owned by the same group count as one voice: their scores are averaged into a single group score, and on each attribute the group's most recent finding speaks for it. Every group weighs the same. There is no hidden ranking of publications.

Normalizing scores

Original scores and scales are stored exactly as published. Numeric scores are converted linearly to 0–100:

normalized = (score − scale minimum) ÷ (scale maximum − scale minimum) × 100, rounded to a whole number.

Examples: 4 out of 5 → 80; 8.5 out of 10 → 85; 85% → 85. Reviews without a numeric score inform the consensus but never receive an invented number. Badges, awards and categorical ratings are not converted unless a specific public rule exists for them.

The Quorum Score

The Quorum Score is the plain average of the group scores of the independent groups that publish a numeric score, rounded to a whole number only at the end (for example 82/100, never 82.347). Groups whose reviews are purely textual count toward the Source Quorum and the consensus, but not toward the number — each product page says how many groups the number rests on. If fewer groups publish a score than the quorum requires, confidence is low. Reviews published after the research cut-off date are left out until the next recalculation.

Methodology changes

• 1.1 (September 2026) — a group's score is the average of all its eligible titles, instead of only its most recent review; the number of groups that publish a score is recorded and shown. Verdicts calculated under 1.0 keep their version until an editor recalculates and approves them.
• 1.0 (September 2026) — first public version.

The quorum rule

No verdict or score is published until all of these hold:

1. the product's identity, variant and region are confirmed;
2. it can actually be bought in the tracked market;
3. there are at least 3 qualifying full reviews;
4. from at least 3 independent editorial groups;
5. no unresolved critical or identity alert affects the evidence;
6. there is a Buyer Signal or Creator Signal for context, unless a documented editorial exception applies;
7. a research cut-off date and the methodology version are set;
8. an editor has validated the eligibility of the sources.

Some categories with slower review cycles may set their own threshold. Thresholds are never lowered retroactively to let a product reach quorum. Before the quorum, a product page explains what is confirmed and what is missing, and shows no score.

Review Consensus and Split Verdicts

Each finding is tied to an attribute configured for the product's category (a chair is not judged on camera quality). For an attribute to be discussed at all, at least 2 independent groups must have commented on it. Then:

• Review Consensus: at least 75% of those groups were positive (a strength) or negative (a common criticism);
• Split Verdict: below that threshold, with both positive and negative findings — shown with the sources on each side.

An attribute score appears only when sources themselves scored that attribute.

Quorum Confidence

• Low: the minimum quorum, a documented exception, or wide disagreement.
• Medium: at least 4 independent groups and a score spread (standard deviation) of at most 15 points.
• High: at least 5 groups, spread of at most 8 points, most reviews from the last 18 months, at least 75% of the category's decisive attributes covered, diversity of countries or methods, no split on a decisive attribute and no open volatility alert.

Agreement between a few sources, or between dependent sources, never produces high confidence on its own.

Human review before publication

Recalculations only create drafts. An editor publishes a verdict after confirming this checklist:

• Identidade, variante e disponibilidade confirmadas
• Mínimo de reviews e de grupos independentes cumprido
• Duplicatas, republicações e primeiras impressões fora do score
• Notas originais e normalizadas conferidas
• Nenhum review textual recebeu nota inventada
• Buyer Signal e Creator Signal separados do score
• Conteúdo patrocinado ou incentivado identificado
• Consensos proporcionais à contagem real
• Divergências materiais visíveis
• Nenhuma linguagem de teste próprio
• Região, data de corte e versão da metodologia aparecem
• Preços e estoques com data
• Fontes realmente usadas listadas

(The checklist is kept in the language of our editorial team.)

Monitoring, corrections and downgrades

New qualifying reviews, corrections, firmware changes, recalls or a methodology change trigger a recalculation. If a published verdict loses its basis, it is flagged and can be suspended; the history is never deleted. Material changes appear in each page's update history. See the corrections policy.

Structured data

Our pages use Product, BreadcrumbList, ItemList, Organization and WebSite markup. We deliberately do not mark verdicts as Review or AggregateRating: a synthesis of other publications' reviews is not our own review, and presenting it as one would misrepresent it.

Limitations

• We depend on what publications review; niche products may never reach a quorum.
• Equal weights treat a lab-tested review and a hands-on review the same. We prefer that to an undisclosed hierarchy, and may revise it publicly in a future version.
• Reviewers can test different units, firmware or configurations. We record the variant, but differences can remain.
• Scores on different scales are not always equivalent even after normalization.
• Prices and availability change quickly; offers show the date they were checked.