Audience research
Part of Planning retail media audience research with six populations kept apart
Retail media buyer personas as evidence cards, with the stereotype fields removed
Build retail media buyer personas as dated evidence cards with clear populations, decision events, contrary findings and limits, never fictional certainty.
Retail media buyer personas should be revisable evidence summaries, not invented characters. A useful card describes a verified role, a recent decision, the evidence behind it and the limits on inference. It does not add a name, portrait, salary, personality or motivation to make incomplete research feel real.
Separate the role from the person
Create different cards for the retailer commercial buyer, advertiser decision-maker, agency operator, platform or data user and measurement approver. Consumers and retail staff need separate research summaries because they experience the communication or workflow rather than buy retail media.
IAB UK's category definition distinguishes on-site, off-site and in-store opportunities. Put the exact environment on the card. Evidence from a sponsored product decision cannot quietly become evidence about store screens or off-site activation.
Use an evidence card
Leave every field blank until source material exists:
| Field | Required record |
|---|---|
| Population | observable role and England connection |
| Decision event | accepted, rejected, revised or operated a defined exchange |
| Evidence base | sample, dates, method and source IDs |
| Task and constraint | participant statement or observed step, labelled correctly |
| Decision authority | documented approval or responsibility boundary |
| Contrary evidence | cases that do not fit the working summary |
| Confidence | analyst-owned explanation, not a decorative score |
| Review trigger | new evidence, changed process or expired fieldwork |
Keep participant words, researcher observation and editorial inference in separate columns. A quotation requires the participant's verified wording and approved publication use. This article contains none.
Add a provenance line to every populated field: source ID, fieldwork date, eligible population, analyst and question or observation version. A claim drawn from one retailer must stay attached to that retailer sample. Do not copy it into an advertiser card merely because the two organisations worked on the same campaign.
Remove stereotype fields
Do not infer buying authority from seniority, technical skill from device choice, business size from a postcode, or accessibility needs from a visible characteristic. Collect personal data only when it changes recruitment, participation support or a defined analysis.
The ICO's data minimisation guidance says personal data should be adequate, relevant and limited to the stated purpose. It also explains that opinions should be identified as opinions with enough context to interpret them. A privacy reviewer should therefore challenge demographic colour that serves no research decision.
Use functional access questions, such as the format or support a participant needs. GOV.UK guidance on finding research participants discusses recruiting actual or likely users and avoiding bias. Its examples come from government-service research; they do not establish a private retail-media population or sample quota.
Publish the boundary
Add a header to every card: HYPOTHESIS, NOT A REPRESENTATIVE PERSON. Show the eligible population, sampling frame, invited/responding/analysed denominators, field dates, missingness and conflicts. A pattern found among retailer staff says nothing about advertiser demand unless advertiser evidence supports it separately.
The Code of Practice for Statistics 3.0 describes quality in terms of intended use, appropriate sources, methods and uncertainty. Applying that discipline helps expose weak cards, but it does not give commercial personas official status.
Review cards side by side for contradictions. If agency operators report a clear hand-off but retailer users cannot identify the approval owner, keep both findings and investigate the difference. Do not average the accounts into a smooth story. The unresolved gap may be more useful to the buyer than a polished persona.
Retire a card when its underlying workflow changes, contrary cases overturn it or the source sample no longer matches the decision. If no approved evidence exists, publish the blank template and keep the persona UNKNOWN.