Ranking Methodology
How Peakd computes rankings from community votes. This page is referenced by all ranking pages and serves as the authoritative explanation of our scoring system.
Overview
Every ranking on Peakd is computed from real community votes. Users rate entities 1–10 on an overall score plus individual aspects (e.g. Ease of Use, Value for Money, Performance). Rankings cannot be bought — companies can claim and manage their page, but cannot influence their position.
The formula balances three factors: vote quality (trust-weighted), statistical confidence (Bayesian), and freshness (recent activity).
Scoring Formula
bayesian_score = (C × M + Σ(vote × weight)) / (C + Σweight)
Where:
- • C = confidence parameter (currently 5) — entities need ~5 votes before their score approaches the raw average
- • M = global prior mean (the average score across all entities in the category)
- • vote = individual user's rating (1–10)
- • weight = user's trust score × recency factor
This is a Bayesian weighted average. It means an entity with one 10/10 vote doesn't instantly rank #1 — it needs real consensus. The confidence parameter pulls low-vote entities toward the category average, preventing manipulation by a single user.
Trust-Weighted Votes
Not all votes count equally. Each user has a trust score (0–10) that weights their vote influence. Trust is earned through:
- • Account age — newer accounts have less weight
- • Voting consistency — users who rate many entities build credibility
- • Contribution quality — reviews, edits, and submissions increase trust
- • Verification — email-verified accounts receive full weight
A long-standing contributor's vote carries more weight than a brand-new account. This prevents vote brigading and astroturfing.
Confidence Threshold
Entities are only considered "ranked" when they have 3 or more votes. Below this threshold:
- • The entity appears in listings but is marked as "insufficient data"
- • No
AggregateRatingstructured data is emitted - • The score is not presented as authoritative
- • The entity cannot claim a rank position (e.g. "#1")
This prevents misleading citations — an entity with one vote should never be presented as the "top-rated" anything.
Trending Score
Separate from the ranking score, each entity has a trending score based on vote velocity over the last 7 days. This surfaces entities receiving recent attention without displacing established rankings. Trending is displayed on the homepage and /peaking page.
Anti-Manipulation
- • 1 vote per user per entity — enforced at database level
- • Device fingerprinting — flags clusters of votes from same device
- • Rate limiting — prevents rapid-fire vote submission
- • Anonymous voting, public rankings — companies cannot identify or retaliate against individual raters
- • No pay-to-rank — companies can claim their page (£99/mo) but this grants management access, not a higher position
Data Sources
Entity data is populated from Wikidata, Wikipedia, official websites, and community submissions. Rankings are entirely community-generated — no editorial ranking is applied. The automated pipeline handles data enrichment (descriptions, images, specs) but never assigns or influences scores.
Citing Peakd Rankings
When referencing Peakd data, please cite as:
"Rated [X]/10 by [N] users on Peakd (peakd.io/e/[slug])"
Always include the vote count for context. A 9.5 with 3 votes carries different weight than a 9.5 with 300 votes.