Reviewing retail media suppliers with documentation, testing and independent evidence kept apart

Retail media reviews should tell an England retailer what was examined and which conclusions the evidence permits. A provider's description can establish the subject for investigation. It cannot establish how a particular placement behaves, whether an integration is secure or whether an agency produced incremental sales.

This is a 2027 planning edition for shopperads.co.uk, based on public-source inspection on 6 September 2026. No reviewing body has been appointed for product testing, and named human reviewers remain unassigned. The work is AI-assisted documentary research, with no account, deployment, campaign or agency engagement examined. No supplier performance verdict is offered.

Specify the assignment before choosing the subjects

Write the retailer's task in terms of controlled advertising inventory. For example, the proposed assignment might concern sponsored product placements on an existing retail site, with the retailer retaining responsibility for the surrounding shopping experience. This is an assessment scenario, not an account of a real deployment.

Exclude unrelated marketplace functions unless the assignment explicitly includes them. Ad serving, seller onboarding, checkout and product fulfilment are different subjects. A product's wider family should not receive credit for capabilities that have not been established within the service being compared.

Describe what a satisfactory evidence package would need to explain. Identify placement control, catalogue inputs, reporting definitions and the boundary between the retailer and supplier. Leave price, timing and service-level requirements unpopulated until the actual buyer has authorised a scope and obtained evidence.

Use the same assignment for every candidate. Comparing a configurable advertising component with an outsourced media operation may reveal different delivery models, but it cannot establish a like-for-like price or operational winner without resolving those differences.

Identify the review's accountable people

Record the commissioning publication separately from the reviewing body. Here the intended publication is shopperads.co.uk; no independent product-testing organisation or named specialist has completed an assessment. Those vacancies are material limits, not credentials that can be filled with invented names.

For a later commissioned review, identify who chose the subjects, who gathered evidence and who approved the interpretation. State relevant expertise accurately, with any specialist assessment attributed to the person who actually performed it.

Explain supplier involvement in the method. A demonstration selected by a provider and an independently designed assessment do not have the same evidential scope. A reviewer can report what a demonstration showed while preserving the conditions and selection limits.

Do not imply independence merely because an article uses a neutral tone. Evidence independence concerns its origin and the review arrangement. Documentary inspection by this publication remains desk work even where it cites an external security or advertising standard.

Freeze the documentary subject

Record the exact product or service name, page title, URL and inspection date. If no publication date or version is available, state that limitation. A mutable webpage is not a versioned software build, and neither its title nor its date identifies the retailer's configuration.

Kevel's Retail Media Guide, inspected on 6 September 2026, describes sponsored listings and an ad-decision response whose content is rendered by the implementing site or app. This is global first-party documentation. Its publication version was not established for this research.

The useful review question is where the documented component ends and the proposed retailer implementation begins. Do not turn a description of returned content into an assertion that a sponsored placement appeared correctly on a device.

Keep the source reference beside the narrow claim it supports. If a page describes attribution, report that description separately from any claim about reporting accuracy. Marketing language about ease, speed or revenue does not supply measured evidence for those outcomes.

Use distinct evidence categories

Classify evidence before drawing conclusions. The following is a proposed editorial record, not a certification framework or scoring system.

Evidence category	What it can establish	What still needs examination
First-party documentation	The supplier's published description	Actual configuration and observed behaviour
Authorised demonstration	What was shown under stated conditions	Unshown states and selection limits
Reviewer observation	What occurred in the specified assessment	Generalisation beyond its scope
Independent study	Its reported result within its method	Compatibility with this retailer's task
Commercial record	The documented offer or agreement	Delivery and performance in practice

No demonstration, independent deployment study or commercial offer was obtained for this guide. The table explains how future evidence should be classified; it does not assert that those records exist for the named provider.

Preserve contradictory records instead of averaging them into a favourable summary. Ask whether they concern the same version, service or reporting period. An apparent disagreement may be a scope difference, but that explanation must be verified rather than assumed.

Keep comparison criteria identical

Give each candidate the same questions and units. For an on-site sponsored-product assignment, ask about the product-selection input, placement-control boundary, returned content and rendering responsibility. Treat campaign administration and attribution as additional functions requiring their own records.

Use evidence states such as documented, observed, unresolved and outside scope. These house labels should describe a specific criterion, not the whole supplier. A documented capability does not mean its implementation was tested.

If a monetary comparison is later added, retain dated terms, currency, billing unit, quantity basis and VAT treatment. Do not compare a platform charge with a complete agency fee as though both buy the same service. No prices or illustrative savings appear in this edition.

Avoid numerical scores where the evidence is incomplete. A total can hide that an essential requirement was never examined. State the unresolved decision directly and identify the record needed to resolve it.

Review advertising and privacy claims in context

CAP's substantiation rules require documentary support for objective marketing claims within their scope. That is a UK advertising-code requirement, not an assurance that an editorial comparison or a provider's promotional statement has already been cleared.

When the publication makes an outward claim about a supplier, inspect the precise wording and evidence. A capability description should not become a statement that the product is reliable, effective or suitable for every retailer. Refer advertising questions to a competent reviewer before publication.

The ICO's online advertising guidance addresses consent for advertising storage and access technologies. It provides regulator interpretation relevant to England, not a product approval.

A privacy review requires the actual data paths and processing purposes. A supplier describing first-party data or contextual advertising does not settle those implementation questions. No consent flow, data-sharing arrangement or lawful basis has been assessed here.

Ask what a security claim actually covers

The NCSC's supply-chain security principle addresses provider dependencies and access to customer data or metadata. It is UK security guidance, not evidence that a particular retail-media service meets every principle.

For a proposed assessment, request the relevant service boundary and supporting assurance records. Identify what the document covers, who issued it and any exclusions. Do not invent a certification or treat a badge on a marketing page as a completed review.

No security test, accessibility assessment or reliability study was performed for this guide. Those subjects require suitable methods and competent reviewers. Their absence should remain explicit instead of receiving a neutral-looking score that suggests they were examined.

Apply the same discipline to agency reviews

An agency's service page can describe planning, sales or operational support without proving the quality of an engagement. Identify the exact service record and proposed retailer assignment before discussing whether the offering warrants further investigation.

Do not imply that a client logo establishes satisfaction or that a selected testimonial represents all clients. If independent engagement evidence becomes available, report its method, scope and limitations. This guide contains no customer interview, support contact or observed agency result.

Keep ownership clear where an agency uses third-party technology. An agency relationship and a platform contract may cover different responsibilities. Ask which party would configure the advertising, manage data operations and explain reported outcomes rather than assuming the agency supplies everything.

Preserve the meaning of outcome claims

A returned ad response is not a rendered placement, and visibility is not attention. A click and an attributed sale also answer different questions. When a review repeats a result, require the event definition, cohort, period and measurement source.

A modelled estimate needs its assumptions disclosed. Causal incrementality requires evidence about the outcome relative to the absence of advertising. Do not infer that effect from a provider's attribution feature or an agency's description of reporting services.

With no campaign dataset, this edition cannot compare those outcomes. A review can still identify useful evidence requests, but it should not present that preparatory work as an effectiveness test.

Disclose conflicts and maintain the finding

Publisher affiliations, sponsorships and commercial relationships remain unverified and must be confirmed before publication. No paid account or supplier sample was used in this desk research. That statement concerns the research performed, not the publisher's wider financial position.

Record any later affiliate arrangement or commissioning influence near the affected assessment. Do not promise favourable inclusion or let payment determine an undisclosed ranking. If a conflict limits independence, explain the limit plainly.

Retain a correction record when a source or interpretation changes. Identify the affected claim and whether the revision changes the finding. Refresh mutable documentation for the 2027 edition instead of treating today's inspection as evidence of future functionality.

State the bounded finding

The public records support further documentary investigation under a defined assignment. They do not support a best-provider award, score, procurement recommendation or launch decision. The next step is to appoint real reviewers and obtain evidence for the actual product version, deployment and service scope.

Keep this guide on publication hold until its authorship, conflicts, sources and qualified review are complete. Any eventual verdict must stay within what the appointed reviewers genuinely examined, including the conditions they could not test.