Audience research
Part of Marketing attribution audience research: six populations
Five practices for marketing attribution customer interviews
Use five reproducible practices for marketing attribution customer interviews, covering eligibility, consent, questions, evidence and stopping rules.
What to take away
- Anchor each interview to a recent decision where an attribution output changed, delayed or missed a business answer, and log the event date.
- Publish the recruitment denominatorframe source, the study’s England boundary and counts at every stage.
- Treat agreement to take part and a UK GDPR lawful basis as two separate decisions with named owners.
- Freeze a numbered discussion guide, pilot it, then version any change instead of editing it mid-fieldwork.
- Record whether each note item was observed, supplied, inferred or unknown, and set stop conditions before fieldwork.
Anchoring the session to a recent decision
For marketing attribution customer interviews, ask when an attribution output last changed, delayed or failed to answer a business decision. Record the event date, the participant's role, the platform or tool involved (for example, Google Analytics 4 or Adobe Analytics) and the evidence they could show.
GOV.UK's in-depth interview guidance tells government teams to recruit current or likely users and to plan questions around the research objective. In a commercial study, likely use must rest on a verified task rather than stated interest in analytics.
Publishing the recruitment denominator
Name the frame source and define the England rule before fieldwork begins. For this study, it is a geographic eligibility boundary, not a legal rule: include participants only when the team responsible for the relevant attribution decision is based in England; record UK-wide roles separately. Our Attribution Audience Research Guide for England is useful alongside six distinct recruitment populations: in-house marketing decision-makers; attribution or analytics analysts; channel marketers who use reports; data or engineering staff responsible for tracking; agency or consultancy practitioners; and procurement or finance reviewers. Classify each participant by primary role and note any secondary roles.
Report a count at every stage so a reader can see where the sample narrowed.
Example: a recruitment denominator for a 20-organisation frame
Recruitment denominator
| Stage | Count |
|---|---|
| Organisations in frame | 20 |
| Invitations delivered | 18 |
| Declined at invitation | 5 |
| People screened | 12 |
| Failed screening | 3 |
| Eligible participants | 9 |
| Interviews completed | 7 |
| Withdrawn after screening | 2 |
| Unusable sessions | 0 |
The GOV.UK recruitment guide discusses defining criteria and the risk of recruitment bias. Record the incentive, and name the groups that timing, channel or access format may have excluded.
Separating agreement to participate from data law
Give participants an accessible information sheet covering the purpose, the researcher, recording and observers. Set out data use, recipients and retention in the same document, and cover withdrawal and the complaint route there too.
GOV.UK's informed-consent procedure gives service teams a practical model. A private organisation still needs a named data-protection reviewer to select and document a lawful basis, because agreement to take part and a UK GDPR lawful basis are related but not identical.
Recruitment must not conceal sales. The ICO separates genuine market research from messages that carry promotion or enable later marketing to the same people.
Freezing and piloting the numbered discussion guide
Our attribution buyer personas article treats personas as hypothesis cards with retirement rules, which is a useful model for prompts you intend to reuse.
Number every prompt and fix the order. Start with what happened, then ask why, and only then test a proposed solution. Place optional probes under the question they belong to and log when each was used.
Use the participant’s actual platform rather than assuming one. For example, ask:
- “Tell me about the most recent time you used an attribution report in Google Analytics 4, Adobe Analytics or another platform to make a decision. What did you look at?”
- “Which report or model did you rely on, and what other evidence did you check?”
- “What changed as a result, and what remains uncertain about the attribution output?”
Freezing and piloting the guide
- Write numbered prompts, each tied to one research objective.
- Pilot with someone who matches the eligibility rule.
- Mark ambiguous wording, missed prompts and any leading question.
- Freeze the guide with a version number and date.
- Issue a new version if a prompt changes, and record the date it took effect.
The Government Analysis Function's questionnaire design guidance covers aims, question formats, respondent burden and testing. Its page carries a review notice, so check its current status before you rely on it.
Keeping observation, statement and inference apart
Notes should mark whether an item was observed, supplied by the participant, interpreted by the researcher or left unknown. Obtain the agreed permission before you record anything.
Field note: [observed] participant shared the dashboard; [supplied] "we dropped last-click in March"; [inferred] budget pressure drove the change; [unknown] who approved it.
Set stop conditions before fieldwork. Stop if the invitation turns promotional, eligibility fails or agreement becomes unclear. Stop as well if unexpected sensitive data appears or the participant becomes distressed.
The report needs the guide version and field dates. It also records the interviewer, the incentive and any conflicts.
- Guide version and field dates recorded
- Interviewer, incentive and conflicts listed
- Role coverage and non-response reported
- Negative cases and eligible refusals counted
Silence is not agreement, so report how many eligible people declined.
Use the Attribution Methods Evidence Grid to keep causal language proportionate to the evidence. An interview account can show how a person used an output or made a decision; by itself, it does not establish that the platform caused the business outcome. Label causal explanations as inferences unless separate evidence supports them.
Common questions
What goes in the recruitment denominator?
Name the frame source, state the study’s England eligibility boundary (the relevant attribution-decision team must be based in England), and report a count at each stage. Include withdrawals, unusable sessions and eligible refusals.
Do I still need a lawful basis if participants agree?
Yes. Agreement to take part and a UK GDPR lawful basis are separate decisions, and the basis should be chosen and documented by a named reviewer.
When should an interview stop?
When the invitation turns promotional, eligibility fails or agreement becomes unclear. Stop sooner if unexpected sensitive data appears or the participant becomes distressed.
What must the final report contain?
The guide version and field dates. Also the interviewer, incentive and any conflicts. Add role coverage, non-response and negative cases.
