Juno

How to run qualitative interviews at scale without losing depth

By Juno · Published · Updated

Qualitative research usually loses depth at scale in two places: the interviews, which drift apart or turn shallow as volume grows, and the synthesis, which falls back on counting what came up most. You keep the depth by designing the study around research objectives rather than a question list, and by using an interviewer that adapts to each person while covering the same ground every time.

Why depth usually breaks at scale

With eight interviews, one experienced researcher can hold the whole study in their head. With a hundred or more, the work spreads across more interviewers and many more transcripts, and the same problems tend to appear.

Consistency drifts

Different moderators probe differently, and the same moderator gets tired. Interviews that started out comparable end up covering different ground, so a theme can look strong only because some interviewers asked about it. Locking everything to a script fixes comparability, but it loses the conversation.

Depth gets traded for throughput

At volume, follow-up is the first thing to go. A generic "tell me more" replaces a real probe, and vague answers get accepted because there isn't time to chase them. The interview still produces a transcript, but it reads like a survey with longer answers.

Synthesis turns into counting

Hundreds of transcripts can't be read the way eight can, so analysis slides towards tallying how often a topic came up. That loses the minority view, and it can hide a split. A message that some people love and others reject looks average once it's averaged.

How objectives-led design keeps depth

A study built from research objectives tells the interviewer what it needs to learn, not what to say. Each participant can take a different route through the conversation, and the study still lands the same coverage. Where exact wording matters, such as a positioning statement or a rating, those questions stay fixed and are asked word for word.

Design is also where most bias enters. A leading question gets a sincere answer to a question that already contained its conclusion, and nothing in analysis can recover the exchange that never happened. Checking the design before fieldwork, and rehearsing the interview, matters more at 200 interviews than at eight, because a flaw in the design is repeated in every one.

How adaptive interviewing keeps depth

Adaptive interviewing means choosing the next question from what the person has just said. In practice that means asking for a concrete example when an answer is vague, and following a contradiction instead of moving past it. Pace changes too: where an answer gets interesting, the interviewer slows down. Because the objectives hold steady, adaptation changes the route through each interview, not the research, so the set stays comparable.

Depth also depends on people feeling able to be candid. A participant who feels pressed gives the acceptable answer, or a short one. Letting people take part in their own time, by text or voice, makes the honest answer easier to give.

Keeping depth through synthesis

At scale, synthesis should be organised around what the study needed to learn, not around which words came up most. Two habits matter most. The first is looking for the people who don't fit a theme, because disagreement shows where a finding stops being true. The second is tracing every claim back to the quote and transcript moment behind it, so anyone can check the reading instead of taking it on trust.

What this looks like in practice

These published studies each ran at a scale that is hard to reach with traditional qualitative methods, and each found something a count alone would have missed.

OrganisationWhat the study foundStudy size
Amber ElectricOne positioning message divided battery owners. Those who liked it chose it more than any other option, and those who didn't rejected it more than any other. A score would have put it in the middle. Because each owner explained their reaction in their own words, Amber came away knowing which customers the message is for.100 battery owners
The Arthritis MovementFive problem statements in people's own words, including that arthritis is dismissed as normal ageing and that mental health support is missing. They now anchor the organisation's policy and advocacy work.239 participants
G+S CommunicationsAustralians care about climate action but feel stuck between intention and friction, and words like "trustworthy" no longer persuade: people want proof. The study was published as Climate Sentiment in Australia 2025.200 interviews

Read the full stories at heyjuno.co/stories.

How Juno handles this

Juno turns a plain-language brief into research objectives, not a discussion guide, and shows the assumptions it made so you can confirm or correct them. The design is checked for leading wording and built-in assumptions, and you can rehearse the interview before launch. Simulated interviews are used only for that rehearsal, never as research data.

In each interview, Juno adapts to the participant while working towards the study objectives and asking any word-for-word questions exactly as written. Recruited interviews are checked for participant effort, and weak ones are flagged with the reasoning visible rather than absorbed into the data. Synthesis is organised around your objectives, and every claim links to the quote and transcript moment behind it, so you can question it in a debrief with Juno. Juno's standards are set by qualitative researchers with 25+ years' experience. See how Juno interviews and our methodology standards.

Frequently asked questions

How many interviews do you need for qualitative research at scale? Enough to compare the groups your decision depends on. Larger studies earn their keep when you need to see where a finding stops holding, such as how one segment reacts differently from another.

Can AI-moderated interviews keep the depth of a skilled human moderator? Judge it by the transcripts. Look for follow-up questions that pick up on what the person said, and for requests for a concrete example when an answer was vague. If every interview has the same rhythm, it's working through a list.

How do you keep interviews comparable when every one is different? Hold the objectives and any word-for-word questions steady, and let the route through the conversation vary. Comparability comes from covering the same ground, not from asking identical questions in the same order.

How do you analyse hundreds of qualitative interviews? Organise the analysis around the study's objectives and look actively for the people who disagree. Every claim should trace back to a quote. Counting mentions is a starting point at most.