Planning retail media audience research with six populations kept apart

Retail media audience research should begin with a decision, not a broad profile of shoppers. An England retailer might need to learn why advertisers reject a placement brief, why agency staff cannot reconcile delivery, or why a customer mistakes paid placement for ordinary merchandising. Those are different questions for different populations.

No research was conducted for this guide. It provides a blank plan for an adult-participant study. A buyer must supply the sampling frame, participant evidence and results, then obtain named ethics, privacy, accessibility, safeguarding and statistical review before publication.

Define the retail-media decision

IAB UK's retail media definition covers retailer or marketplace controlled advertising space, data assets and in-store opportunities. That category boundary helps identify who may know something, but it does not make everyone a research audience.

Write one decision sentence:

We need evidence from [eligible population] about [recent task or event] so [named owner] can decide whether to [change, test, retain or stop] [defined retail-media process] by [review date].

The sentence must identify the retailer or marketplace property, media environment, advertiser exchange and England connection. Keep on-site, off-site and in-store journeys apart because they involve different people and records. State what would change the decision and what evidence would stop the study.

Keep six populations separate

The legal retailer or marketplace controls the relevant property or inventory. Its commercial buyer may commission a media service, while operational staff approve products, creative, access and reporting. These staff are not interchangeable.

An advertiser chooses whether to fund a communication. An agency may plan, create or buy it under delegated authority. A platform serves or records activity. A data or measurement supplier handles defined inputs and outputs. Each organisation can have a buyer, user and approver with different evidence.

Consumers and retail staff may be affected by the advert, controls, tracking or destination. They can describe a task they experienced, but they are not evidence of advertiser demand. Likewise, an advertiser interview cannot establish that a customer recognised a paid placement.

Create a population register with these fields:

Population	Decision role	Eligible recent event	Frame owner	Excluded from inference
Retailer commercial buyer	approves the exchange	reviewed the defined proposition	retailer	shopper attitudes
Advertiser buyer	accepts or rejects a brief	made a documented decision	advertiser	retailer capability
Agency operator	manages a delegated hand-off	handled the named campaign stage	agency	authority not delegated
Platform or data user	operates a system step	used the exact workflow version	employer or supplier	product performance outside that step
Affected adult consumer	encounters the journey	completed or abandoned the defined task	ethically recruited frame	advertiser budget
Retail staff member	performs or supports a task	used the process during the study period	retailer	national workforce prevalence

Do not infer residence, authority, disability, income or technical skill from a name, postcode, job title or public profile.

Build a frame that can answer the question

The GOV.UK Service Manual advises researchers to recruit actual or likely users and explains how recruitment choices create bias in finding user research participants. It is written for government services, so a commercial retail-media team should use it as method guidance rather than a rule that fixes sample size.

Record the target population, source list, frame date, contact provenance, eligibility screener and every exclusion. Count eligible, invited, responding, completing and analysed participants separately. A convenience sample of existing clients can describe those clients; it cannot represent all England advertisers. Staff nominated by a manager may also withhold criticism, so record who introduced each participant and offer a private response route.

England must be an observable rule. It might mean a person performs the relevant job for an England retail operation or experienced the specified placement in England. A company registered in England does not prove that its respondent worked on the defined event.

Recruit without disguising a sales approach

Research participation, use of personal information and later marketing are separate decisions. The ICO says genuine market research does not count as direct marketing, but a survey that contains promotion or collects details for future campaigns becomes direct marketing in its electronic and telephone marketing guidance. Do not append a sales pitch or move research contacts into a marketing list.

Give each person an accessible information sheet that names the controller, research purpose, activities, observers, recording, recipients, retention, publication plan, contact route and withdrawal process. GOV.UK guidance on informed consent for user research treats participation as voluntary and requires participants to understand what will happen and how their data will be used.

Participation consent is not automatically the UK GDPR lawful basis. The ICO's research principles and grounds guidance makes that distinction. A privacy specialist must document the actual basis, purpose and any additional condition before collection.

Make the instrument answerable

Use a recent, specific event. Ask an advertiser buyer to reconstruct the last in-scope brief they accepted, rejected or returned for clarification. Ask a retailer operator to show the sequence of evidence fields with synthetic records. Ask an affected consumer what they believed was paid and what action they thought a price or control would take.

Put neutral factual questions before opinions. Record the exact question version, order, interviewer prompt and materials shown. Separate what the participant said, what the researcher observed, the analyst's interpretation and what remains unknown. Do not insert a real campaign, customer order or account into a research prototype unless the owner has specifically authorised that use.

The Government Social Research Profession's ethical assurance guidance identifies sound conduct, specific informed consent, enabling participation and minimising harm as separate principles. Its remit is government research, but those questions make a useful ethics review for a commercial project.

Plan accessibility, incentives and safeguarding

Ask about participation adjustments without requiring a diagnosis. Offer usable information and consent formats, remote or in-person routes, breaks and communication support where needed. The government guide to running sessions with disabled people says researchers should speak to the participant directly and avoid assumptions about what they can do or feel.

Record any incentive before recruitment, who receives it, when it is given and whether it could exert pressure. It pays for time; it does not buy a favourable answer or quotation.

This plan excludes children by default. If the question genuinely requires children or adults at risk of harm, stop. Assign a safeguarding lead, obtain a project-specific ethics route, define responsible-adult involvement, safe disclosure handling and withdrawal, and use age-appropriate materials. Do not infer vulnerability from shopping or campaign data.

Control the research data

The ICO's research provisions guidance is under review following legal changes. It says its examples are not complete compliance cases. Treat it as a current, qualified source and recheck it before fieldwork and publication.

Create a data map for recruitment details, screeners, notes, recordings, transcripts, incentives and analysis outputs. For each item, record purpose, source, access, recipient, retention, deletion and correction route. Keep contact details away from response data where possible. The GOV.UK guide to managing research data and participant privacy recommends collecting the minimum recruitment information and anonymising public extracts.

If a survey or transcription supplier is proposed, record its exact legal entity, service, data role, subprocessors, transfers, access controls, retention and export. A privacy page is evidence of a statement, not proof of the buyer's configuration. NCSC supplier assurance questions provide a starting point for access, incident, data and recovery evidence.

Analyse without manufacturing prevalence

Prepare the analysis plan before reading answers. Define codes, who applies them, how disagreements are resolved and which responses are excluded. Preserve missing answers and withdrawals rather than treating them as negative responses. For every proportion, show the numerator, denominator, population, field dates and uncertainty.

Interview themes explain the analysed sample. They do not establish how common a view is across England. A survey also needs a defensible frame and response assessment before anyone calls it representative. The UK Statistics Authority's Code of Practice for Statistics 3.0 provides current quality principles for suitable sources, transparent methods, limitations and uncertainty. Applying those principles voluntarily does not turn commercial research into official statistics.

Ask participants to verify any quotation and context proposed for publication. Prefer paraphrased, anonymised findings when a distinctive role, retailer or incident could identify someone. Keep a correction log and remove data when the approved withdrawal or deletion route requires it.

Make a bounded decision

The final evidence table should show the decision question, eligible population, frame coverage, invitations, responses, exclusions, missingness, finding, contrary evidence and confidence. The owner may proceed to a synthetic test, revise the proposition, commission more research or stop.

Do not merge samples to create a larger-looking audience. Retailer staff, advertiser buyers and consumers answer different questions. Keep the page on publication hold until the named research, ethics, privacy, accessibility, safeguarding and statistical reviewers can trace every public statement back to an approved source and an appropriate population.