How to review branding agencies: declared evidence and like-for-like criteria

Treat supplier review as a sequence of evidence decisions, not a contest for a universal winner. Begin with the purchase question, then follow the branch that matches the evidence available. At each point, record what supports a finding, what remains unresolved, and what must be checked before an appointment can be justified.

If the purchase question is still unsettled

If the review cannot say what decision it is meant to support, pause the supplier search and define that decision first. Without a clear boundary, providers may be judged against different assumptions, while an impressive portfolio can distract from whether the proposed work is needed or suitable.

Write down the intended audience, outputs, channels, territory, languages, delivery dependencies and decision date. Compare the proposed commission with a manual alternative, such as retaining the current identity and correcting selected assets. Use the commission planning questions to make the choice explicit before assessing agencies.

A public service project for a regulated organisation and a naming brief do not raise identical needs. Visual examples do not, by themselves, establish research design, rights transfer, accessible delivery, governance or exit arrangements. Leave each unanswered matter open until evidence that relates to that particular need is available.

Rule of thumb: settle the purchase question and its alternative before comparing providers.

If the name on the proposal does not settle responsibility

If the public-facing trading name does not identify who accepts contractual responsibility, ask the candidate to name the legal entity that will contract, invoice and receive data. A familiar name or matching company address is a useful lead, but it is not proof of responsibility or work quality.

Request the full legal name, company number, registered and trading addresses, VAT details where relevant, parent or guarantor, subcontractors, and the party that will invoice and receive data. Confirm the proposed signatory's authority, then reconcile the named entity across the proposal, invoice and bank details. Filed details can be checked through the official company search.

The examples locate Wolff Olins Limited, company 01945130, Pentagram Design Limited, company 01599748, and Ragged Edge Design Limited, company 06328903. These are identity leads, not a census or ranking of English providers. The published review examples do not turn a company record into approval of an agency, its work or its contract.

Companies House status is one input, not a quality finding. Its information has not been verified or validated by Companies House, and a record does not approve an agency, contract or work. Save the source and access date; treat an address match as a prompt for further checks, not a legal conclusion.

Rule of thumb: confirm the entity taking responsibility, not merely the name used in public.

If candidates describe different commissions

If proposals cover different services, do not treat their claims as equivalent. Define one service unit and ask each provider the same question about it. Where the work remains unclear, record equivalence as unresolved rather than turning a list of capabilities into a comparable offer.

Translate each proposal into the same buyer problem, activities, outputs, delivery people, rights, privacy, security, accessibility, implementation, governance, exit and billing basis. Include assumptions, exclusions and VAT treatment, then use like-for-like comparison criteria for each specific offer.

Wolff Olins describes services including insights and measurement, purpose and strategy, identity and activation. Pentagram describes multidisciplinary work across identity, strategy, products, digital experiences and communications. Neither description establishes equivalent contracted scope, staffing, price or delivery standard. Check the relevant terms rather than inferring them from a provider's general description.

Rule of thumb: compare the work a supplier would agree to deliver, not the breadth of its public claims.

If one requirement must stand on its own

If a requirement is decisive, assess it as a separate gate. Do not let strength elsewhere offset a failure or an unknown. Keep the criterion, evidence and outcome visible so that a strong creative impression cannot conceal a material concern.

Mark each test pass, fail, conditional or unknown, and record its criterion and evidence. An unknown should not become a midpoint in a weighted total just to produce a tidy result. If a score is used, disclose its scale, evidence rule, weight and formula alongside the underlying evidence.

For rights, list names, concepts, artwork, type, imagery, code, research and third-party materials. Put ownership and permitted use in writing, then ask qualified UK IP reviewers to examine title, licences and assignment wording. A copyright notice may help identify a source, but checking the actual terms remains necessary.

For accessibility and equality, specify the people, artefacts, channels, standard and test method. An inclusive-design claim is not an accessibility report. State whether web content is included and which accessibility standard is being applied; public-sector requirements may not settle every question for a particular engagement.

For privacy, distinguish research, service administration, analytics and marketing, and consult the data protection review to establish actual privacy roles rather than relying on a contract label. For security and continuity, examine access, systems, locations, incident handling, supplier governance and exit. NCSC supplier assurance questions can inform checks but do not certify an agency.

For commercial terms, compare currency, VAT treatment, billing unit, period, assumptions and included work. Leave a price unknown when none is published. Review acceptance, changes, cancellation, liability, confidentiality, data terms, subcontracting, correction, handover, deletion and continuity as distinct matters rather than folding them into a general impression.

Rule of thumb: keep critical requirements separate; creative strength cannot cancel an unresolved gate.

If a source seems to prove more than it says

If sources address different questions, preserve their limits instead of combining them into one proof. Separate official records, provider statements, independent material and buyer-supplied documents. Then check whether each item speaks directly to the criterion being assessed.

Use Companies House records for filed company information while retaining their stated limits. An award, certification, directory entry or membership establishes only its disclosed scope and current status; the absence of an adverse decision is not an approval. Keep a source note with the access date and the wording relied upon.

Treat a case study as the provider's account of a project. It can help frame a reference-check question, but it is not independent proof of performance. The data protection principles can frame a query, but evidence is still needed about the actual research, administration, analytics or marketing roles involved.

Rule of thumb: match a source to the narrow claim it can support, and leave wider claims open.

If the evidence comes only from public pages

If no candidate has supplied procurement evidence, describe the work as desk research and limit conclusions to that scope. State what was examined and omitted, who selected the material, relevant interests, and how a reader can raise a correction or challenge.

The desk research closed on 5 September 2026. No agency was contacted, paid, commissioned or given advance notice. No client reference, proposal, contract, accessibility report, security evidence, pricing schedule or working file was obtained.

Disclose hands-on work and untested areas, funding and conflicts, the scoring formula or an explicit no-score rule, and the correction, challenge and update process. State whether affiliate links, referral fees or supplier-funded placements exist. These details allow readers to distinguish first-hand experience from checking publicly available material.

A desk review can record pages and dates, confirm that records exist and compare stated fields. It cannot establish collaboration quality, strategic judgement, file quality, accessibility in use, response time, delivery capacity or business effect. Mark each omitted test beside the conclusion rather than implying that it took place.

Rule of thumb: declare the research boundary and interests before stating findings.

If a hands-on question remains unanswered

If the question concerns collaboration, strategic judgement, file quality, accessibility, response time, delivery capacity or business effect, public pages cannot resolve it. A review method can organise public evidence, but its limits should remain beside the verdict.

Before appointment, use authorised synthetic material for a short working exercise if procurement permits it. State who created the fixture, what the exercise measures, how unpaid speculative work is avoided, who owns the output and how each candidate receives equivalent instructions. Do not add personal or confidential information merely to make the exercise seem realistic.

Record the artefacts inspected and each path left untested. If no hands-on exercise took place, say so beside the conclusion rather than burying it in a footer. A presentation or public portfolio cannot replace an exercise designed to answer the specific question under consideration.

Rule of thumb: state the test, its purpose and what it cannot assess.

Applying the criteria to the three examples

Here is the same set of gates applied at desk level to the three agencies named above. The result is a statement of what the public record supports, not a verdict.

Like-for-like in this review means the same fields asked of each agency. On those fields, the three differ only in their described scope, and the remaining gates cannot be compared until proposals, references and terms are obtained. To run the method yourself, send the same one-page brief and the same questions to each shortlisted agency.

If important evidence is still missing

If material matters remain unanswered, issue a bounded finding rather than naming a universally best agency. The next step may be a defined check, conditional progress tied to specified evidence, a hold, or exclusion against a documented criterion.

For the public-record examples, the supported finding is that three London trading and company records were located and selected service descriptions can be compared at desk level. Suitability remains unassessed: proposals, references, named people, terms and evidence for the gates are still needed before appointment.

Recheck company records, first-party pages, conflicts and regulatory sources on publication day, and make changed facts visible when corrected. Have the method and wording checked by a qualified reviewer before you rely on a published comparison.

Rule of thumb: report only what the evidence supports, and name the next check when the record is incomplete.