Comparing retail media suppliers with the evaluation unit held constant

A retail media supplier comparison needs a common purchasing unit. Here that unit is a retailer-side service supporting paid on-site inventory on one specified England retail website. The comparison concerns documented scope, not measured performance. Wider audience services, other properties and off-site delivery require separate specifications before their costs can be compared.

Records and comparison limits

The official records below were accessed on 6 September 2026. They are mutable global product pages, with no reliably established publication date used here. Each supplier receives the same questions about on-site function, integration boundary and reporting. None was contacted or tested, and the sources do not establish an England-specific offer.

Service record	Documented on-site role	Evidence still needed for the shared scope
CitrusAd retailer platform	Sponsored products and retailer-controlled ad serving	Included integration work and the exact reporting specification
Criteo Commerce Yield	On-site monetisation and supply-side reporting	Contracted inventory configuration and retailer responsibilities
Kevel Ad Server	Sponsored listings with management, decision and reporting APIs	Supplied components versus retailer-built workflow and rendering

These statements describe what the vendors document, not equivalent packages. The table is deliberately non-ranked. A broader product description should not be interpreted as a better result, and an unanswered commercial question is not evidence that the supplier lacks the function.

Hold the evaluation unit constant

Specify the same website, placement format and catalogue scope for each response. Use a single defined test task, such as creating an eligible sponsored product campaign and reconciling its recorded events. This is a proposed acceptance task, not a test conducted for the article.

Ask every supplier to identify who performs each step. Keep the same event definitions even when product vocabulary differs. Compare ad decisions with ad decisions, rendered impressions with rendered impressions and attributed purchases with purchases using the same attribution definition.

Do not treat an ad request as a viewable impression or a platform-attributed order as proof of incremental sales. Before measuring anything, document the evidence required for the specific term. This editorial comparison method makes no claim that the named systems calculate metrics identically.

Request comparable commercial units

Set one quotation period and ask for GBP amounts with VAT treatment stated. Separate the initial implementation fee, recurring service charge, usage-based charges and optional support. Require the supplier to identify the billable event, minimum commitments and excluded services.

Where rates use different units, retain the original rate beside any scenario calculation. Use only verified contractual quantities and clearly label assumptions. No prices, volume forecasts or total-cost results were collected here, so there is no numerical cost league table.

Include retailer engineering and operations effort as an unpriced requirement until the responsible teams estimate it. Ask whether the proposed support covers the same hours and incident types. A quote for a different responsibility split cannot establish a like-for-like saving.

Compare assurance evidence separately

NCSC guidance on cloud audit information addresses the records customers need to understand service activity. Ask all candidates for the same evidence about access, changes and relevant alerts. Have a security assessor judge the response rather than assigning unsupported safety scores.

Keep privacy and commercial review outside the product-feature table. Confirm the actual legal entities, processing roles and any international access with qualified UK advisers. This comparison is neither legal advice nor a statement that a documented feature makes the service compliant.

Record a decision without inventing a winner

For each mandatory row, retain the official record, supplier clarification and later acceptance result separately. State which evidence is missing and what would resolve it. Recheck product documentation before publication or contracting; the valid outcome of this desk comparison is an enquiry shortlist, not a purchase recommendation.