Why the retail media benchmark register should stay empty until the definitions match

Retail media benchmark research needs comparable campaigns and definitions before it can supply an England performance reference. The public material inspected for this article did not establish a compatible England dataset for the requested measures. Benchmark values therefore remain UNKNOWN, with no estimate substituted for missing evidence.

What this research examined

The inspection date is 6 September 2026. The scope is retailer or marketplace advertising inventory, with England results sought separately from UK or international totals. Public primary-source searches and direct inspection covered measurement standards, a UK advertising-spend release and government evaluation guidance.

Candidate evidence needed a defined event, denominator, cohort, reporting period, time zone, attribution window, exclusions, correction treatment, direct source and uncertainty. Counts could mark denominator not applicable; non-attribution measures could do the same for the window. An undisclosed field was treated as a compatibility question rather than silently assumed.

This was a bounded desk search, not a systematic census of every commercial dataset. Private campaign records and paid datasets were not obtained. Failure to establish a benchmark here does not prove that no suitable evidence exists elsewhere.

Separate a reported study from a performance inference

IAB UK's Digital Adspend 2025 release, published in March 2026, reports UK advertising investment research conducted with Oliver Wyman. It includes retail media within market spending. No monetary figure from it is reused as a campaign benchmark here. Only the public release was inspected; the member-only full report was not accessed.

Publisher inference: a UK market-spend release cannot by itself establish an England campaign click-through, visibility or incremental-sales benchmark. Its geography and unit answer a different question. This is a compatibility judgement about the inspected release, not a finding that the underlying research is inaccurate.

The IAB Europe May 2026 measurement standards provide industry definitions. They are not an observed England campaign cohort. A recommended measurement convention should not be relabelled an achieved performance level.

Keep the benchmark register unpopulated

The following fields are research targets, not market estimates. Every row inherits the complete evidence requirements above.

Candidate benchmark	Value	Definition still needed
Click-through ratio	UNKNOWN	Qualifying clicks and matched eligible impressions
Viewability ratio	UNKNOWN	Qualifying viewable events and measured eligible base
Attributed-order ratio	UNKNOWN	Credited orders, eligible denominator and allocation window
Incremental outcome	UNKNOWN	Outcome unit, analysis population and causal design

Keep ad requests and served or returned responses outside a rendered-placement denominator unless the adopted definition explicitly justifies their role. Rendering, viewability and attention remain different observations; no attention benchmark was established by this research.

For each eventual candidate dataset, record format, product scope and collection coverage. Preserve original currency for any monetary result and document a GBP conversion only if a valid comparison requires it. Do not introduce an exchange rate, price or performance value without supporting evidence.

Assess whether the proposed comparison answers the question

HM Treasury's evaluation methods annex, updated in May 2026, distinguishes impact designs and their limitations. It offers methodological context for evaluating causal evidence, not a retail-media benchmark series.

Label a study's reported result as reported, and identify any publisher calculation or inference separately. Keep modelled estimates distinct from observed outcomes. A vendor-selected campaign example requires its selection limits to remain visible; it should not automatically become a reference population for England retailers.

State the next evidence request

Request a source dataset and methodology matching the intended metric, geography and cohort. Reassess compatibility before calculating or ranking anything, and preserve corrections if the source changes. Until that evidence is available, UNKNOWN is the defensible register entry. Publication also awaits real named authors, qualified measurement review and confirmation of publisher conflicts.