Reviews
Part of PR agency reviews in England: five named checks, from Companies House to the CMA's fake-review rules
How to read a PR agency review in England: what each kind of evidence proves
Judge a PR agency review methodology in England by its evidence: independent facts versus first-party claims, references, performed tests and expiry.
A PR agency review is only as good as the evidence behind each verdict. Credentials decks, coverage reports and testimonials are written by the agency itself. A buyer in England can test any review, or the buyer's own comparison, by asking what every source proves and what it cannot. Once the evidence holds up, the selection checklist turns it into an appointment decision.
What to take away
- Register and regulator material proves narrow facts only, not capability or suitability.
- Proposals, portfolios, terms, case studies and testimonials are first-party claims.
- A reference needs its selector, the service and the period discussed on record.
- A demonstration viewed by a buyer is desk evidence, not a performed test.
- A missing mandatory record is HOLD, not a low score, and a review expires when the entity, service or terms change.
Sort the evidence into two layers
Label every source as independent evidence for a narrow fact or as a first-party claim. Register and regulator material is independent only for the narrow fact it proves. Proposals, portfolios, terms, case studies and testimonials are first-party claims, however polished they look.
| Kind of evidence | Layer | What it can show | What it cannot show |
|---|---|---|---|
| Companies House filing | Independent, narrow | A bounded identity check | Solvency, capability or suitability |
| ICO register entry | Regulator material, narrow | That the organisation appears on the register | Whether its data handling is lawful |
| PRCA or CIPR membership | Professional body statement | That the agency accepted the body's code | Skill, results or fit for the buyer's task |
| Credentials deck, proposal, terms | First-party claim | What the agency says it does | That the claim is true or comparable |
| Coverage report, case study | First-party claim | What was published and what the agency chose to show | That the agency caused the coverage, or a result any buyer should expect |
| Permissioned reference | Selected evidence | One client's view of one service | A result any buyer should expect |
| Buyer's own test | Primary evidence | Whether a task can be done in the buyer's environment | Performance on other tasks |
Public relations in England is largely self-regulated, so no regulator vouches for an agency's competence. The PRCA and the CIPR are professional bodies with codes of conduct, and the PRCA also runs its Consultancy Management Standard. Membership or accreditation shows that the agency accepted a body's rules or assessment, not that it can do the buyer's task.
Companies House's disclaimer says publication does not mean company-supplied information was verified or validated. A filing therefore supports bounded identity checks, not solvency, capability or suitability. Directories, awards, anonymous testimonials and historic case studies without a comparable denominator carry no weight as evidence.
A coverage book lists what was published, in which outlet and when, but editorial decisions stay with the outlet, so it cannot show that the agency controlled the result. A report that totals advertising value equivalents rests on a measure the Barcelona Principles reject. The measurement guide explains why.
Read references as selected evidence
A reference is one client's account, and someone chose that client. Ask who selected each referee, and which service and period the referee discussed. Get permission for the buyer's questions and for any later publication of the answers.
Record the questions asked, the answers that are missing and any conflict of interest. A reference with gaps in that record shows the agency's choice of voice, not a representative result. Put the same questions to ask before hiring to every referee so the answers can be compared.
Ask for a performed test, not a demonstration
A demonstration viewed by a buyer is desk evidence, not a performed test. A feasible buyer test is one task the buyer can run in an authorised environment, with the expected evidence fixed before the test starts.
For each test, keep a short record:
- Tester and date.
- Authorised environment, with synthetic or permissioned fixtures.
- Script and expected evidence.
- Observed evidence, any exception and the reset.
Suitable tasks include correcting an approved fact, suppressing a contact, withdrawing an artefact such as a draft release, recovering buyer-owned access and exporting the handover pack.
Security claims need the same care. The NCSC says Cyber Essentials certification gives organisations a way to gain assurance that suppliers have implemented fundamental technical controls. That is a narrow fact about basic controls, and it does not prove an agency's security overall.
Keep criteria separate and scores blank
Apply the same evidence fields to every candidate and keep each criterion separate. Quality evidence cannot offset a failed privacy, rights or security review. A missing mandatory record is HOLD, not a low numerical score.
Leave buyer weights, thresholds and scores blank until the buyer has agreed them. The CAP Code's misleading-advertising section supports documentary substantiation where an objective advertising claim falls within scope, so a claim in a first-party document needs its proof attached.
Data roles follow the same rule. The ICO's guide to controllers and processors covers how to determine whether an organisation is a controller or a processor, so judge roles from real decisions, not contract labels. Map privacy and PECR purposes before comparing agencies.
Check the review's own limits
A trustworthy review states its subject, service version, evidence cut-off, untested areas and who may rely on it. It also says whether the outcome is admit for testing, conditional, reject for a defined failure or HOLD for missing evidence.
Look for disclosure of funding, affiliate relationships, gifts, free access and client engagements, and for the named people who checked the facts. A disclosure policy that records funding, roles and AI use in one place makes that check quick.
Expire the record when the entity, service, subcontractors, material terms, data route, evidence owner or buyer requirement changes. Until a dossier is complete and a buyer test is feasible, HOLD is the honest verdict, as on a shortlist built on a strict admission rule.
Before you act
- Label each source as independent evidence or a first-party claim.
- Record who selected each referee, plus the service and period discussed.
- Fix the expected evidence for each test before it starts.
- Leave buyer weights, thresholds and scores blank.
- Name the subject, service version and evidence cut-off.
Common questions
What does a Companies House filing prove?
It supports bounded identity checks only. Companies House says publication does not mean company-supplied information was verified or validated, so a filing does not prove solvency, capability or suitability.
Can a strong case study offset a failed security check?
No. A case study is a first-party claim, and strong results elsewhere do not cancel a failed security check. A missing mandatory record is HOLD.
When should a review record expire?
Expire it when the entity, service, subcontractors, material terms, data route, evidence owner or buyer requirement changes. Recheck every source on the day the record is used.
