1. Market eligibility
We only build a report where the market is big enough to compare fairly. A market that falls short is turned down before any payment, not stretched to fit.
1.1 How a market is defined
A market is one category in one area: the businesses in your category within a 3 to 5 mile radius of your postcode. The full competitor list appears on page 03 of every report, so you can see exactly who was compared.
1.2 The minimum we will build on
At least 12 comparable businesses, each with at least 10 reviews. Below that, we tell you before charging and suggest a wider radius or refund any deposit. Roughly one in twenty requested markets falls below the threshold.
1.3 What a typical market looks like
| Market type | Businesses | Reviews |
|---|---|---|
| Single-postcode small market | 12–25 | 150–500 |
| Mid-density urban category | 25–60 | 500–1,500 |
| Dense regional category | 60–120 | 1,500–4,000 |
2. Data collection
Every report is built from one public source, read exactly as any visitor would read it.
2.1 One public source
Every figure comes from public reviews on Google Maps. We do not sign in as a Google user, collect anything that needs a login, or pay for elevated API access. What we read is what any unauthenticated visitor to the same page would see.
2.2 What we collect for each business
- Name, address, category, and the phone number and website where they are public.
- The average star rating and total review count, as Google displays them.
- Every public review with its text, its 1–5 star rating, its posted date, and its language.
- The business’s reply to a review, where one was posted, with its date.
2.3 What we never collect
Reviewer personal data beyond the display name shown on the review: never an email address, a phone number, or any private contact. Nothing behind a login, and no private business dashboards.
3. Review integrity
Three kinds of review are kept out of all or part of the analysis: suspected fakes, reviews Google has removed, and translations we are not confident in.
3.1 Suspected fake reviews
We exclude reviews from accounts with no other public activity and a single review on the business in question, where the pattern fits a known cluster of suspected fakes. Every excluded review is noted in the methodology page of the delivered report.
3.2 Reviews Google removes
A review flagged or removed by Google between collection and publication is excluded, even when our snapshot was taken before the removal.
3.3 Reviews in other languages
Non-English reviews are machine-translated for analysis, with the original preserved. Where translation confidence is low, the review is left out of theme analysis and the exclusion is noted.
4. Analysis checks
Every figure is computed from the reviews themselves, and theme tagging gets a human spot-check on a 10% sample of every report.
4.1 Figures are calculated, not estimated
Ratings, volume, distribution, sentiment per business and per aspect, emotion counts, the competitor benchmark, top phrases and the competitive map are all computed directly from the reviews. Sentiment is scored positive, neutral or negative at both the review and the aspect level.
4.2 Themes are checked by a person
Each review is tagged against a fixed taxonomy of themes built for the category, from a structured read of every review rather than a keyword count. A person spot-checks the tagging on a 10% sample of every report.
4.3 Like is compared with like
Review volume varies sharply by category — restaurants attract far more reviews per customer than accountants — so we never compare across categories. Every comparison in a report is inside one market.
5. AI-written sections
A language model writes the narrative half of every report from calculated figures, and every report says which half is which.
5.1 What a language model writes
A report has two halves. The calculated half is drawn directly from the reviews. The written half — the executive summary, the market narrative, sentiment themes, the persona, the SWOT, the strategic priorities and the recommended actions — is written by a language model from a representative sample of review text plus the calculated figures. Every report’s appendix says so in the same words.
5.2 When a section is left out
Any written section may be omitted when the data is insufficient or the generation fails. When that happens, the report ships without it.
5.3 The persona is labelled for what it is
The synthesised customer persona is not a real person, a forecast, or a statistical aggregate. The 0–10 recommendation score it gives is the persona’s own answer, not a measured survey result and not a claim about your real customers.
6. Freshness
Every report is one dated snapshot, typically 2 to 4 days old when delivered, and is never refreshed afterwards.
6.1 One dated snapshot
Each report is built on a snapshot taken at the start of the work, and the cover page records its date. We typically deliver within 2 working days of collection, so the data is 2 to 4 days old when you first read it.
6.2 No silent refreshes
A delivered report is never updated in place. If you want to see how a market has moved a year later, that is a new report.
7. Limits we disclose
Public review data has known biases, and the methodology page of every delivered report flags them in the same language:
- Reviews skew to the extremes. People review when they are delighted or annoyed, so the middle is under-represented. The distribution analysis is framed with that in mind.
- Volume varies by category. Which is why we compare only within one market.
- Translation is imperfect. Low-confidence translations are kept out of theme analysis, and noted.
The written Strategy section is a structured starting point for an internal conversation, grounded in the data. It is not bespoke advice and not a substitute for your judgement.
8. Independence
We read reviews; we do not influence them, and we have no tie to any review platform.
8.1 No platform ties
Rate My Ratings is independent and not affiliated with Google or any review platform.
8.2 Research, not reputation management
We do not respond to reviews, flag reviews for removal, or contact your customers. A report tells you what the market says; what you do with it is up to you.
8.3 Data stays with the report
We do not sell data to third parties. The reviews collected for a report are used to produce it, and are then retained only so we can answer questions about it.
9. Non-negotiables
These 5 commitments are non-negotiable:
- We never contact reviewers, under any circumstance.
- We never write fake reviews, suppress real ones, or advise on either.
- We never sell review data on to third parties.
- We never respond to reviews on a business’s behalf.
- We never run reputation-management or review-removal services.
The full method behind every figure is on our methodology page.
10. Accountability
Rate My Ratings is run by Deventral Ltd, a company registered in England and Wales and incorporated on 4 November 2022, with its registered office at Office 408, Screenworks 22 Highbury Grove, London, N5 2ER. The company’s record is public at Companies House under company number 14461518, and the DEVENTRAL name has been a registered trade mark with the UK Intellectual Property Office since 5 July 2024, as UK00004037110. Both can be checked on the registers that hold them.
If you have a question about a report or about these standards, email [email protected] and we will reply within 1 working day, Monday to Friday, UK time. Anything to do with personal data is covered in our privacy notice.