Methodology · public record

The rules behind the ranks.

The Beauty 100 combines an editorial product assessment with a confidence-adjusted community signal. It is designed to reward durable value without letting a small, coordinated, or commercially connected vote become the list.

01

The one public question

Would you buy it again at today’s price?

A star can mean almost anything. A repeat purchase forces performance and price into the same decision. Members answer yes or no, identify their relationship to the product, and can add moderated context.

Would you buy it again at today’s price?

YesNo
Answers submitted after publication inform a later edition—not the rank you are viewing.
02

Eligibility first

A product earns the right to be scored.

Editors assess the product version against a fixed rubric before community data is allowed to influence placement. A version must score at least 70 out of 100 on the editorial scorecard. Products below that gate do not enter the ranking, regardless of popularity or campaign volume.

  • Version-specific.A materially changed formula is evaluated as a new version; old evidence does not silently transfer.
  • Price-aware.The list shows a dated price snapshot because the same performance can represent different value at a different price.
  • Category-aware.A cleanser and a fragrance are judged in their own use context before competing in the global list.
  • No paid entry.Sampling, sponsorship, affiliate availability, or a badge conversation cannot make a product eligible.
03

Three inputs, one controlled signal

Editorial quality leads. Evidence earns influence.

E

Editorial assessment

A fixed rubric establishes the product’s quality and value case.

C

Community signal

Eligible yes/no answers are adjusted for sample confidence rather than read as a raw popularity poll.

F

Freshness

A small recency input prevents a once-great product version from coasting indefinitely.

Editorial scorecard · 100 points

30 Performance20 Usability20 Value15 Evidence10 Distinctiveness05 Transparency
Why confidence adjustment matters

A product with two yes answers is not treated as more certain than a product with hundreds. Low-sample results stay closer to a neutral prior and remain more editorial-led; community influence grows only as eligible evidence accumulates and is capped so it cannot overwhelm the rubric.

Published calculation

Category prior
(category weighted yes + 10) ÷ (category effective n + 20)
Product community rate
C = 100 × (product weighted yes + 20 × category prior) ÷ (effective n + 20)
Confidence
effective n ÷ (effective n + 40)
Community weight
30% × confidence
Editorial weight
95% − community weight
Final score
E × editorial weight + C × community weight + F × 5%

E, C, and F are expressed on a 0–100 scale. Community influence starts at zero, grows with effective evidence, and can never exceed 30%. Freshness is always 5%. There is no abrupt minimum-vote cliff.

Rank stability: when final scores are within 0.75 points, the prior order is retained to prevent noise-driven churn. Remaining ties resolve by final score, editorial score, then product ID for deterministic output.

04

Disclosure is part of the data

Every answer names its relationship.

Relationship is required for votes and context notes. Commercially connected responses remain visible for disclosure and abuse review, but do not receive the same ranking treatment as an independent purchase.

Purchased
The member paid for the product.
1.00×
Personal gift
It came from a friend, family member, or another personal relationship.
0.75×
PR sample
A brand or agency sent the product without a purchase.
0.50×
Brand / retailer relationship
The member has a material professional or commercial connection.
Not disclosed
The answer is retained with reduced influence because independence cannot be assessed.
0.50×
05

A published list is a record

Ranks freeze when an edition is published.

The page displays a dated snapshot, not a live stock ticker. New answers can change the evidence used in a later recalculation, but cannot rewrite an already published rank. Recalculation is a deliberate admin action that writes an immutable revision. There is no automatic daily cadence: a reviewed run is deliberately published, with movement measured against the prior snapshot.

  1. 1Evaluate

    Editors score eligible product versions using the current rubric.

  2. 2Recalculate

    Eligible community evidence and freshness are applied under the published scoring rules.

  3. 3Review

    Integrity checks look for version drift, coordinated activity, missing price context, and moderation issues.

  4. 4Publish

    The edition, ranks, product versions, and price date become a frozen public snapshot.

06

Context, with guardrails

Notes are moderated, not manufactured.

Submitted notes are plain text, require relationship and use context, and enter a moderation queue before publication. Links and HTML are not accepted. Members can delete their own published notes and report spam, hidden relationships, harassment, impersonation, privacy concerns, or medical misinformation.

Moderation can publish, reject, or remove a note for policy reasons. It does not rewrite a member’s opinion into house copy.

07

The commercial wall

The rank is not inventory.

A brand cannot purchase eligibility, score, placement, movement, a positive note, or suppression of a negative answer. Affiliate links can generate commission after the ranking decision; they do not enter the score.

For brands

Our transparent roadmap separates future badge licensing, aggregate insight work, panels, events, and sponsorship from the editorial rank.

Read the commercial policy

Questions or corrections

Methodology should be inspectable.

If a product version, price snapshot, relationship label, or published note appears wrong, tell us what should be checked. A correction changes the record transparently; it is not a private route to a better rank.

Contact the editorial team