Audience research
Paid social agencies buyer personas: coded records to cards
Code records into one evidence card for a paid social buyer persona: card fields, a hypothetical worked card, second-analyst checks and expiry triggers.
A paid social buyer persona is useful only as the last step of analysis. Permitted interview notes and briefs are coded, contrary cases are kept, and one card states what the records support. Each card names a decision, a role and a record count. It carries no stock photograph, invented quotation, salary, budget or motive.
What to take away
- Code permitted records before writing a card. A card summarises codes, not impressions.
- Name a decision and a role on the card, never a fictional character.
- Keep contrary cases and missing populations beside the pattern they contradict.
- Treat a small sample as evidence of how a decision happens, not of how common it is.
- Give every card an expiry trigger, and leave the persona slot empty until evidence exists.
Write the codebook before reading the records
Write the codebook before analysis, so the codes come from the research question and not from the first vivid transcript. Give each theme a definition, an inclusion rule and an example drawn from permitted notes. Themes can cover the task, the platform, the approval path and the constraint.
Have a second analyst code a sample and log every disagreement. If one commissioning owner wants direct platform billing while another cannot support it, preserve both conditions rather than averaging them into a vague preference. The log shows where a code definition is too loose, and the research protocol decides which populations the records come from.
Code the task, not the title. Titles are prompts, not proof of who decides, so record who performs each task. Commissioning a campaign, releasing funds, granting platform access, approving creative and checking measurement are separate tasks. Do not merge them because one interviewee mentions them all, and define each decision role once.
Fill the evidence card field by field
An evidence card has eight fields, and each one points back to coded records. A field the records cannot support is marked as not established, never completed with a plausible guess.
| Field | Where the entry comes from | If the records are silent |
|---|---|---|
| Decision and role | Coded task statements | Leave the card unwritten |
| Organisation unit and geography | The screening record and the study's stated scope | State the nations and regions covered |
| Number of records | A count of coded records behind the pattern | Enter the actual count, not an estimate |
| Method and field dates | The interview, brief or debrief log | Mark the evidence as desk research |
| Recent task, platform, inputs, approval path | The most recent task described in the records | Write not established |
| Constraints and variation | Coded constraints and differences between accounts | Write not established |
| Contrary cases and missing populations | The disagreement log and the screening gaps | Write none recorded |
| Confidence label and next test | The analyst's reading of the counts | Pick the lower label |
Use a stable code such as budget-approver / multi-site retailer / England-head-office only for evidence scoped to England. For UK-wide work, record the relevant nation and region rather than treating England as the whole UK. Small samples can reveal how a decision happens, but not how common the pattern is in the population studied.
A hypothetical worked card for a Meta campaign decision
This is a hypothetical example, not a research finding or a claim about typical buyers. Suppose a client interview and brief record this path: a paid social manager plans in Meta Ads Manager, a performance marketing manager approves the spend, and the agency's media buyer implements it.
The account director relays changes. The card names the decision and the roles, the organisation unit and recorded UK geography, and Meta Ads Manager as the platform. It names the interview and brief as the method, with the actual record count and field dates entered from the source before use. Unsupported fields are marked as not established.
Note what the card leaves out. It does not say how often this approval path occurs, and it gives no budget. Both would need records that the example does not have, so the card marks them as unknown. A hypothetical card demonstrates the format. It is not evidence.
Work with thin evidence and label it
Small agencies without primary research can use permitted briefs, debriefs, CRM records and public ad material as limited desk evidence. Meta Ad Library can show advertising material, not internal budgets or decision paths. IAB UK research can provide market context, not proof of a particular client's buying behaviour.
Label the source on the card and keep the resulting persona provisional. Evidence gathered while researching rival agencies describes suppliers, not buyers, so it should not fill a buyer card. If the evidence comes from one organisation or a convenience sample, say so on the card.
Keep identity and permissions off the card
A card carries no name, portrait or quotation that a participant did not supply. Where a combination of job, sector and region could reveal a person, do not call the card anonymous. Store contact details apart from coded findings, and set deletion dates for transcripts.
A participant may agree to an interview without agreeing to a public quotation or sales contact. The ICO's valid consent guidance explains that agreement to participate in research should not be confused with consent used as a UK GDPR lawful basis. A named privacy professional, such as the Data Protection Officer where one is appointed, decides the actual basis.
The article on data protection across a campaign sets out which party holds which duty.
Promotional reuse is a separate decision. The CMA collection on reviews and social media endorsements says reviews and online content must be genuine and accurate and endorsements made clear, but it does not authorise reuse of participant words. Obtain separate permission for the wording and its context before it appears in paid social advertising.
Check the CAP Code and the advertising rules for the claim itself, as a separate check from the permission.
Apply a decision test before anyone uses the card
Use a card only when its underlying records are current, the represented role matches the decision and contrary evidence is visible. Add an expiry trigger for a platform, policy or buying-process change, such as a change to the way Meta campaigns are bought or approved.
Before a card influences targeting, strategy or sales, have named roles review it:
- A research ethics reviewer checks the method and the potential for harm.
- A DPO or privacy lead checks the lawful basis and identifiability.
- A statistician or research analyst checks the strength of the evidence.
- An account director or other commercial approver checks relevance to the offer.
These are agency safeguards, not a universal statutory approval panel. Until the evidence exists, keep the persona slot empty rather than filling it with plausible fiction.
Before you act
- Write the codebook with definitions, inclusion rules and examples.
- Have a second analyst code a sample and log disagreements.
- Fill each card field from records, or mark it as not established.
- Label convenience samples, desk research and single-organisation evidence.
- Keep contact details apart from coded findings.
- Add an expiry trigger and name the reviewers before the card circulates.
Common questions
What does a card record that a character sketch does not?
It records the decision and role, organisation unit and UK geography, record count, method and field dates. It also holds the recent task, platform, inputs and approval path, constraints, contrary cases, a confidence label and the next test. It has no fictional name or budget.
Can a card quote a participant?
Only with separate permission covering the wording and its intended context. Agreeing to an interview is not agreement to a public quotation, and a quotation must not mislead or hide its commercial context.
When should a card be retired?
When its expiry trigger fires, such as a change to a platform, a policy or the buying process, or when its records are no longer current. A card built from one organisation, a convenience sample or desk research alone should never be presented as a market-wide persona.
