Five moves that make creative-tool customer interviews worth the hour

Good interviews uncover how advertising work happens now. They do not ask people to endorse a planned feature or predict a purchase.

Inclusion method: the five practices below were selected because each changes the reliability or safety of a creative-workflow study and has a live public-sector or regulator record. Research date: 5 September 2026. Geography: the intended users are in England, while the cited government material applies across UK public-service research. Exclusions: no recruitment firm, interview platform or participant incentive is compared, and no interview result is reported.

What to take away

1. Recruit people who face the decision

Choose current or likely users of the workflow under study. For client approval, recruit people who send, review and resolve feedback, not a generic marketer panel.

The government's participant guidance calls for including people with access needs and warns against recruitment bias. Its typical sample guidance covers public-service rounds, not a universal commercial minimum.

Screen on recent behaviour, decision authority and working context. Avoid collecting the client's name, unreleased campaign details or protected characteristics unless the question genuinely requires them and the data route has been reviewed.

2. Ask for the last real example

Open with a neutral prompt such as: "Tell me about the last advert you had to adapt for another channel." Then ask what happened, which artefacts moved between people, where a decision stalled and how the final version was recognised. GOV.UK's page on in-depth interviews recommends open questions, real examples and a tested discussion guide.

Do not ask, "Would automated resizing save you time?" That supplies the desired answer and skips the current workaround. A future feature can be evaluated later in a separate prototype task.

3. Observe the work without taking client material

Ask a participant to reconstruct the sequence with cleared or synthetic assets. Note handoffs, version names, approval cues and moments when they leave one application for another. The guidance on taking research notes distinguishes observed actions from a researcher's interpretation.

If a participant begins exposing a live account, customer list or confidential brief, stop the task. A realistic session does not require the research team to copy material it has no authority to hold.

4. Make consent specific and reversible

Before notes or recording begin, explain who is responsible for the data, why it is collected, who will receive it, the retention period and how to withdraw. The Service Manual's informed-consent guide says participants should know whether observers are present and whether the session is recorded.

Keep evidence of the consent with a participant code, while storing contact details separately. If transcription or another supplier processes the recording, name that arrangement in the information provided to the participant.

5. Analyse while the evidence is fresh

Hold a structured review soon after the round. Record observations first, group related evidence second, then agree findings and next actions. The official analysis method recommends involving observers and keeping observation distinct from interpretation.

A single vivid comment is not a segment finding. Preserve disagreement and negative cases. End with a decision record stating what changed, what remains uncertain and which claim needs another round. That makes interviews useful without turning a small qualitative study into a market statistic.

Before closing the work, audit the sample as well as the themes. Note failed recruitment, early withdrawals, sessions affected by technical problems and groups absent from the round. Those limits belong beside the decision because a polished summary can otherwise make incomplete coverage look deliberate.

Before you act

Common questions

What should I avoid asking in a creative-tool customer interview?

Do not ask questions like would automated resizing save you time, because that supplies the desired answer and skips the current workaround. A future feature can be evaluated later in a separate prototype task. Instead, open with a neutral prompt about the last real example and ask what happened.

How should I handle consent and data during these interviews?

Before notes or recording begin, explain who is responsible for the data, why it is collected, who will receive it, the retention period and how to withdraw. Keep evidence of consent with a participant code, store contact details separately, and name any transcription or supplier arrangement in the information provided.

What should I do if a participant exposes confidential client material?

Stop the task straight away. A realistic session does not mean your team should copy material it has no authority to hold.