Food influencers: customer interviews without guesswork

What to take away

A London restaurant group running three campaigns a quarter with food influencers can learn more from six recorded customer interviews than from a month of dashboard watching.
Recruit interviewees before you brief creators, not after, so the questions you ask are not shaped by the campaign you are already committed to.
Six to ten interviews per audience segment is a workable stopping point for a small operator, judged by whether new themes stop appearing rather than by a fixed number.
Consent for recording and for using quotes sits with the ICO's direct marketing guidance, not with the creator's contract.
Score every interview against the same rubric so one enthusiastic diner does not outweigh four lukewarm ones.

The order below is a sequence, not a ranking. Apply it to any UK nation. Data protection and electronic marketing are UK-wide matters: the ICO enforces the UK GDPR, the Data Protection Act 2018 and PECR. Advertising claims and disclosure fall to the Advertising Standards Authority and its UK codes. Food-safety matters are enforced by local authority environmental health services in England, Wales and Northern Ireland, and by Food Standards Scotland in Scotland, so check the relevant body before you publish findings that touch those areas.

1. Define the audience segment

Pick one segment per round: people who booked a table after seeing a creator post, or people who saw the post and did not book. Mixing both groups in one session produces contradictions you cannot resolve. If your diners are not segmented yet, our guide Food creator partnerships: target audience breakdown gives you a starting set of groups to choose from. Useful starting segments include diners who booked after seeing a creator post, diners who saw a post but did not book, regulars who ignore creator content, and lapsed diners who booked after a creator post more than 90 days ago.

2. Set the inclusion criteria

Write down who qualifies before you recruit. A Manchester operator, for example, might require: booked within the last 90 days, arrived via a tracked creator link, spent above £30 a head, and not employed by the venue. Criteria stated in advance stop you from quietly interviewing only the people who already like you.

Inclusion criteria to set in advance

Booked within the last 90 days
Arrived via a tracked creator link
Spent above £30 a head
Not employed by the venue

3. Choose the geography

Decide whether you are studying one city, one nation or the whole of the UK. A Bristol-only sample cannot support claims about Cardiff. Label every finding with its geography, so nobody reads a local result as a national one.

4. Recruit without incentives that skew answers

A free dessert is a reasonable thank-you and a poor recruitment tool, because it attracts people who want the dessert. A modest voucher issued after the interview is cleaner. For invitations to past customers, decide whether the message is direct marketing and check PECR as well as the UK GDPR and Data Protection Act 2018; the ICO's step-by-step guide to direct marketing for your small business is relevant if it is.

5. Ask about behaviour, not approval

Ask what happened, not whether they liked it. "Walk me through the last time a food creator's post made you book somewhere" produces usable detail. "Do you trust food influencers?" produces a polite yes. Keep questions under 30 words each, and avoid stacking three asks into one sentence.

Use this core script with every participant. Open with: “I’m interested in what you remember; there are no right answers.” Ask:

What do you remember seeing in the creator’s post?
Where did you see it, and what caught your attention?
What did you do after seeing it?
Did it affect whether, where or when you booked? How?
What made the post feel trustworthy or untrustworthy?
If you did not book, what stopped you?
What information, if any, would have helped you decide?

Question types: behaviour vs approval

Ask about behaviour

Example: Walk me through the last booking
Output: Usable detail
Length: Under 30 words

Avoid approval questions

Example: Do you trust food influencers?
Output: Polite yes
Length: Avoid stacked asks

6. Record and transcribe with consent

State plainly what you will record, how long you will keep it and who will read it. Get separate written permission for recording and for each kind of quotation, and give people an easy way to withdraw. Store recordings separately from booking data, so a marketing list and a research file do not merge by accident.

Choose and document a lawful basis under the UK GDPR and Data Protection Act 2018 for processing personal data. Check whether PECR applies before sending electronic invitations; permission to take part in research does not automatically authorise marketing.

Use wording like this, filling in the brackets: “I agree to take part in [interview] for [purpose]. I agree to audio recording: [yes/no]. I agree to an anonymised quote: [yes/no]. I agree to a quote attributed to me: [yes/no]. My data will be kept for [period] and accessed by [people]. I can withdraw by contacting [contact] until [point].”

7. Run the session consistently

Same opening, same core questions, same closing across every interview. Two interviewers should not diverge, or your themes will reflect the interviewer rather than the audience. Keep sessions to around 30 minutes, and leave the last five for anything the interviewee raises unprompted.

8. Score against a shared rubric

Score each interview on the same dimensions. The rubric below keeps scoring comparable across interviewers. A total of 15 or more suggests a strong signal worth acting on; 8 to 14 suggests a theme worth watching; below 8 means you need more interviews in that segment.

Dimension	Score	What a high score looks like
Recall	one to five	The diner describes specific details from the creator's post without prompting.
Trust	one to five	The diner explains why the post felt credible, or names the specific doubt that stopped them.
Booking intent	one to five	The diner describes a change in what they did, or would do, about booking.
Barriers	one to five	The diner names a concrete obstacle—timing, price, location or occasion—rather than a vague reason.

Total the four scores. Use the same scale in every interview.

Scoring rubric for each interview

Dimension

Recall: Specific post and venue, unprompted
Behaviour change: Booked, visited or recommended
Trust: Names what made creator credible
Objection: Clear reason for not acting

What a high score looks like

Recall
Behaviour change
Trust
Objection

9. Turn findings into a brief

Convert themes into a one-page brief: audience and geography; campaign or post; what the audience noticed; what stopped them; what they did next; anonymised evidence; limitations; and the next test. Give that page to creators before briefing them.

Worked example (fictional): Manchester restaurant “Market Kitchen” briefs food creator “City Bites” to show its weekday menu. Interview diners who saw the post but did not book: ask what they recall, what stopped them and what would change their decision. These names are placeholders, not a real campaign.

Where a creator's paid format is part of the plan, TikTok's advertising products for branding set out the disclosure and format options available. If a partnership looks anti-competitive, for example an exclusivity clause that locks rival venues out of a whole city, the process for reporting a market problem to the CMA explains how to raise it.

For the wider method behind these steps, our guide Food creator partnerships: audience research guide sets out sampling, timing and how to keep interview findings comparable across quarters.

Common questions

These are the questions restaurant teams ask most often before a first round of customer interviews.

How many customer interviews are enough?

For a single venue or small group, six to ten per segment is usually workable. Stop when two consecutive interviews add no new themes, and record that stopping point so the next round is comparable.

Should creators sit in on the interviews?

Generally no. Their presence changes what diners are willing to say about creator content. Share anonymised findings with them afterwards instead.

Do we need consent to quote a diner?

Yes, and separately from consent to record. Ask again before publishing a quote, even a short one, and confirm whether they want their name, first name only, or nothing at all.

Can we run interviews for Scotland and England together?

You can, but report the two separately if any rule or market condition differs. A combined average hides exactly the differences you are paying to find.