What a review proves

A review proves that a person, at a date, wrote a statement on a platform. It may also prove, if the platform verified it, that the person had a transaction or employment relationship with the company. It does not prove that the content is true. These are three different facts and a dossier keeps them apart.

OpenDossier therefore records feedback as third-party claims with the platform, the date range, the number of reviews, the platform's verification method and the licence under which the data was obtained. Individual review text is not reproduced as fact; aggregate patterns are described, and the company's reply is shown.

Used this way, feedback answers narrow questions well: whether customers report billing or delivery problems at a rate that has changed over time; whether former employees describe the same governance issues that a litigation record hints at; whether a company has a pattern of responding to complaints. It answers no question about legal liability, financial health or security posture.

Sampling bias and small numbers

Reviews are self-selected. People with strong experiences, good or bad, write more than people with ordinary ones, producing the familiar distribution heavy at one and five stars. Platforms that invite every customer after a transaction reduce but do not remove the bias; platforms that rely on organic submission amplify it. Employer reviews skew toward leavers.

Small samples make this worse. Ten reviews over five years tell you almost nothing, and a single one-star review can move an average by a point. The dossier states the count and the period for every feedback claim and does not report an average at all below a sensible minimum. Where a platform shows a ratings distribution, the distribution is more informative than the mean.

Comparisons are made only like for like: a company's rating against its own history on the same platform, or against companies of similar size and sector on the same platform with the same invitation method. Cross-platform averages are not computed.

Licensed data and verification of the relationship

Scraping review sites breaches most platforms' terms and produces data whose provenance cannot be shown. OpenDossier uses licensed review data where it is available, for example Trustpilot's developer APIs, which provide business-profile and review data under contract, with the platform's own flags for invited versus organic reviews and for reviews the platform has removed or flagged.

Licensed data lets the dossier say how a review was collected: whether the reviewer was invited after a verified transaction, whether the platform matched the reviewer to a customer record, and whether the business responded. These are facts about the relationship, not about the content, and the dossier labels them that way.

Where no licensed source exists for a platform, the dossier may note that public reviews exist on that platform, with a count and date observed, and leave interpretation to the reader. Employer-review platforms are handled the same way, with the additional care described in the next section.

Small teams, anonymity and fairness to individuals

Employee feedback about a company of 15 people is, in practice, feedback about identifiable individuals. A review that criticises the CTO at a small company names a person. Reproducing it in a dossier risks unfairness and, in some jurisdictions, defamation or privacy claims, and it may expose the reviewer.

OpenDossier does not quote individual employee reviews. For companies below a headcount threshold it reports only that reviews exist, their count and the period; above the threshold it reports aggregate themes with counts. Statements about named or readily identifiable individuals are excluded from aggregation. The same restraint applies to customer reviews that name staff.

Platforms themselves apply minimum sample sizes before showing ratings; the dossier respects those and does not reconstruct what a platform withholds. Where an engagement genuinely turns on culture or management conduct, the right route is reference calls with the company's consent, recorded as company-facilitated and attributed accordingly.

Right of reply

Every feedback claim in a dossier is visible to the company once it has claimed its profile, and the company may reply through the Trust Workspace. The reply is dated, attributed to a named officer and shown alongside the claim. It does not remove the claim; a reader sees both.

Replies are most useful when they add evidence: a changed billing process with the date, a platform's confirmation that a batch of reviews was removed as fraudulent, a resolved complaint. Replies that merely dispute are shown too, labelled as the company's position.

The same mechanism applies to litigation, regulatory and press records. It serves two purposes: fairness to the company, and accuracy for the reader, because the reply often contains the fact that explains the record. A company that engages with its file tends to produce a better file for everyone.

The FTC rule and the UK regime

In the United States the Federal Trade Commission's Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465) took effect on 21 October 2024. It prohibits, among other things, fake or AI-generated reviews by people without experience of the product, buying positive or negative reviews, undisclosed insider reviews, company-controlled review sites presented as independent, suppressing reviews by threats, and misrepresenting that a displayed set of reviews is representative. Civil penalties apply per violation; the amount is adjusted annually (check the authority).

In the United Kingdom the Digital Markets, Competition and Consumers Act 2024 added fake reviews and the publication of reviews without reasonable steps to prevent fakes to the list of banned practices, enforced by the Competition and Markets Authority with direct fining powers.

For a dossier this has two consequences. Feedback data is used only in ways consistent with these rules, and a counterparty's own review practices are themselves a diligence point: a company found to have bought or suppressed reviews has a regulatory record, which goes in the legal section with its source.