Trustbadgely
Find a brand

Methodology · Trust Score v1

How we score

The Trust Score is a number from 0 to 100. The formula is public, every input is linked on the brand's page, and brands can ask us to correct mistakes.
See it on a brandRequest a correction
SourcesChecks

Ratings we read

Google ratingsPlaces API · 4.3 from 12,480
App StorePublic feed · 4.7 from 210,000
Brand site ratingsschema.org · 4.6 from 8,900

Checked by the agent

Returns policyPolicy page · 30-day window
Company registryCompanies House · Active
TrustpilotLink only · licence pending
GooglePolicyRegistry

Only data we may use

Official APIs, public feeds, the brand's own pages and company registries. No scraping, no copied reviews, no reviewer names.

Weighted by log10(ratings + 1)

Google4.312,480
App Store4.7210,000
Brand site4.68,900
Penalties−0
Trust ScoreHigh confidence91

Weighted by how many people rated

Each rating becomes 0–100 and is weighted by the log of its rating count, so 50,000 ratings count more than 50, but no single source drowns out the rest.

Confidence on every score

Online-only brands can have little public data. Every score says whether we're highly, moderately or barely confident in it.

1. What goes in

We only use data we are licensed or permitted to use:

  • Google ratings through the official Places API. For chains with stores we use the head office or flagship location plus a rolled-up average, labelled “in-store reviews”.
  • App Store and Google Play ratings from public feeds and official listing data.
  • Ratings on the brand's own site, read from schema.org markup, respecting robots.txt.
  • Review platforms such as Trustpilot, only through licensed access. Until then we link out and show no score.
  • Public signals: returns policy, contact options and company registry status.

We store numbers, dates and links. We don't copy review text or store reviewer names.

2. The formula

Each source rating is converted to a 0–100 scale. Sources are averaged, weighted by how many ratings they have, so a source with 50,000 ratings counts more than one with 50:

base   = Σ (rating / 5 × 100) × log10(count + 1)
         ───────────────────────────────────────
                 Σ log10(count + 1)

score  = base − penalties   (kept between 0 and 100)

3. Penalties

  • No visible contact details: −10
  • No visible returns policy: −10
  • Inactive company registration: −20

4. Confidence

Some brands, especially online-only ones, have little public data. Every score carries a confidence level:

  • High: 3 or more rating sources and at least 1,000 ratings in total.
  • Medium: 2 or more sources, or at least 200 ratings.
  • Low: anything less. Read these scores with care.

5. AI summaries and the agent

Summaries are written by AI from the stored data above and refreshed when a brand's data changes. The agent answers only from that data, cites a source for every claim, and says so when it doesn't know. We describe what customers report; we don't state claims about a brand as fact.

6. What we don't do

We don't verify individual third-party reviews, and brands cannot pay to change their score. Every figure shows its source and an “as of” date.

7. Corrections

If something is wrong, send a correction or removal request. We aim to respond within 72 hours.

Check any brand

Find a brand