Retail media risk scenarios with triggers and responses but no invented incidents

Retail media risk scenarios should describe what to inspect if a defined change occurs. They are not forecasts of incidents. For England retailers controlling advertising inventory, the scenarios below identify triggers, evidence, possible consequences and responses without likelihood scores or estimated losses.

These are publisher planning scenarios prepared on 8 September 2026 from public guidance. No retailer incident, campaign fault or AI test was observed. Their order follows operating topics rather than ranking severity, probability or commercial importance.

A proposed change in advertising data use

Trigger: a new feature or recipient changes the relevant advertising data path. Evidence needed: the actual purposes, device activity, recipients and responsibility map. Possible consequence: an unresolved question about permission or processing scope. Proposed response: refer the arrangement for qualified privacy review before relying on it.

Evidence class: current regulator guidance. The ICO's online advertising page addresses consent for advertising storage and access technologies. It does not establish that any retailer has breached a rule or that a hypothetical feature will be unlawful.

Keep the scenario tied to the actual change. A new tool name alone does not prove a new data flow, and a familiar supplier does not settle the assessment without evidence.

An unsupported AI-produced advertising statement

Trigger: a reviewer identifies a discrepancy between a proposed advertising claim and its supporting record. Evidence needed: the wording, source, relevant output and review finding. Possible consequence: unsupported promotional content. Proposed response: withhold the affected claim, correct the evidence or wording and investigate the discrepancy.

Evidence class: current advertising code. CAP's substantiation requirements concern documentary evidence for objective marketing claims. The scenario assumes no error frequency, AI capability or automated reliability level.

Human review is a proposed house control, not proof that every legal or accuracy question has been resolved. The responsible reviewer needs authority to reject the claim; no actual review outcome is reported here.

A change in measurement definitions

Trigger: an adopted event or attribution specification is revised. Evidence needed: the earlier and revised definitions, scoped exports and correction record. Possible consequence: results become incomparable. Proposed response: explain the definition break before using the series for a performance claim.

Evidence class: current industry methodology. The IAB Europe May 2026 standards provide media and attribution concepts, not a prediction that a particular retailer will change its reporting.

Preserve requests, returned responses, rendered placements and viewability separately from attention. Clicks and credited sales do not establish causal incrementality. A modelled estimate needs assumptions disclosed; no hypothetical reconciliation result is invented.

A changed supplier dependency

Trigger: a proposed service change introduces or alters an external dependency. Evidence needed: the service map, data access, assurance scope and responsible parties. Possible consequence: a gap in the retailer's understanding of the arrangement. Proposed response: obtain the missing assurance and assign the relevant decision to a competent owner.

Evidence class: UK security guidance. The NCSC's supply-chain security principle addresses such dependencies. It does not establish a breach, outage probability or universal service-restoration time.

Do not infer an incident merely because assurance is incomplete. Record the unresolved question and the evidence needed to distinguish uncertainty from an observed failure.

A useful exercise ends with an evidence request that a named colleague can answer. Avoid recording a scenario as resolved merely because it has been discussed; keep the unanswered operational question visible.

Maintain the scenario record

For each applicable scenario, identify the actual owner and update trigger. Preserve the distinction between an announced plan, current documentation and an observed change. A planned release should not enter an incident record as though it had already occurred.

No likelihood, financial impact or staffing requirement has been calculated. Review the scenarios when the relevant operation changes and at publication. Genuine authorship, specialist review, confirmed publisher conflicts and current sources remain necessary; this desk guide makes no launch, legal-compliance or security verdict.