Semi-Structured Interview in Qualitative Research: A Practical Guide

Mariam's fourth interview ran 20 minutes over. The participant, a night-shift warehouse supervisor, drifted away from question six and spent a quarter of an hour on a workaround nobody had mentioned. Mariam let her. That digression became the central theme of the thesis. Interview seven was the opposite: she stuck to the guide, asked all ten questions in order, and got ten short, polite answers she never quoted. (Mariam is a composite, not a real person.)

Both interviews used the same guide. The difference was what she did with it.

That is the whole craft of the semi-structured interview in qualitative research: a guide firm enough that every participant covers the same ground, and loose enough that the supervisor can take you somewhere you didn't plan to go. This guide covers what semi-structured means, when to choose it, how to build and run the guide, consent and recording, and the hour after. Two assets are built in: a guide template you can copy and a probe bank for the moments when an answer is thin.

It's written for graduate students, UX researchers and market researchers with a study to run. If you want the recording to become a transcript draft and a summary while you write your memo, try ParrotNotes free on your next interview.

What a semi-structured interview in qualitative research is

A semi-structured interview is a one-to-one conversation organized around a short list of open questions prepared in advance, in which the interviewer follows up on what the participant says rather than reading a script. DeJonckheere and Vaughn, in an open-access methods paper in Family Medicine and Community Health, describe it as "a dialogue between researcher and participant, guided by a flexible interview protocol" and supplemented by follow-up questions, probes and comments (DeJonckheere and Vaughn, 2019).

The "semi" sits between two other formats.

The semi-structured interview keeps the same core topics for everyone while letting each person answer in their own order and their own words, which is why it is the default in so many master's theses, UX discovery projects and market research studies.

When the semi-structured interview is the right method

Choose it when your research question asks about experience, reasoning or meaning. DeJonckheere and Vaughn note that qualitative research questions "typically start with 'What', 'How' or 'Why'", and that a meaningful project can be done "with as few as 8–12 participants" when the questions are tight and the sample is well chosen.

It is the wrong method when you want to count how often something happens (use a survey) or see what people do rather than what they say (observe them). One more test: can you record? A semi-structured interview without a recording means a notebook, and a notebook means you miss the follow-up.

Designing the semi-structured interview guide

Kallio and colleagues reviewed ten methodological papers on guide development and set out a five-phase framework: identify the prerequisites for using semi-structured interviews, retrieve and use prior knowledge, formulate the preliminary guide, pilot-test it, and present the complete guide (Kallio et al., 2016, Journal of Advanced Nursing). In practice that comes down to four habits.

Start from the research question, not from questions. Write it at the top of the page, then the two to four things you must be able to say at the end of the study. Every interview question has to serve one of those. If it doesn't, cut it.

Build three layers. DeJonckheere and Vaughn describe a guide as "a short list of 'guiding' questions" with three kinds of question in it:

• An opening "grand tour" question. Easy, broad, and about the participant's own world: "Tell me about a typical shift." It sets context and settles nerves.
• Core questions. "Five to 10 questions" that together answer the research question. Each one open, neutral and clear.
• Planned and unplanned follow-ups. The planned ones sit under each core question. The unplanned ones come from the probe bank below.

Fix the wording. Their rule: "All questions should be open ended, neutral, clear and avoid leading language." Start questions with "Tell me about" or "Walk me through", which asks for a story instead of a verdict, so "Do you use the export?" becomes "Tell me about the last time you exported something." Save "why" for later, split double-barreled questions, and swap your literature's jargon for your participants' words.

Pilot it twice. Once with a colleague for timing, once with a real participant. You'll find the question nobody understands, the one that takes 12 minutes, and the one that always gets answered under question two. DeJonckheere and Vaughn are blunt: "it is important to pilot test the interview."

The interview guide template

Copy this into a document, replace the bracketed text, and keep it to one page. One page is a guide; three pages is a script you will read.

INTERVIEW GUIDE: [study title]            Version: [n] / [date]
Research question: [one sentence]
I need to be able to say: 1. [ ]  2. [ ]  3. [ ]
Target length: 45 min (hard stop 60)

OPENING (3 to 5 min)
- Thanks, purpose in two sentences, how the data will be used
- Recording consent: ask, wait for a clear yes, confirm on the recording
- Grand tour: "Tell me about [a typical day / your role / the project]."

CORE QUESTIONS (30 to 40 min)
Q1 [topic 1]: "Tell me about the last time [event]."
   Planned follow-ups: What happened first? Who else was involved?
   Serves: aim 1
Q2 [topic 2]: "Walk me through what you did when [situation]."
   Planned follow-ups: What did you expect? What surprised you?
   Serves: aim 2
   (five to 10 core questions, each with one or two planned follow-ups)

CLOSING (3 to 5 min)
- "Is there anything I should have asked and didn't?"
- "Is there anyone else you think I should talk to?"
- Next steps, thanks, stop recording

AFTER (10 min)
- Memo: what stood out, what surprised me, what to change in the guide

The "Serves" line is the discipline. If a question doesn't serve an aim, it's curiosity, and curiosity goes in the follow-ups, not the core.

The probe bank

Probes are the "semi" in semi-structured. The guide gets you into a topic; the probe gets you the detail you can quote. DeJonckheere and Vaughn list six kinds (wait time, echo, verbal agreement, expansion, explanation and a request for reasoning). The bank below puts phrasing on five of them and adds three of its own. Print it on the back of the guide.

One rule: never probe a yes/no answer with another yes/no question.

Running a semi-structured interview

Tom runs discovery interviews for a logistics software team. His first ten ran 90 minutes because he asked every question, every time, and nothing after minute 60 made it into the findings deck. (Tom is a composite, not a real person.)

Keep it under an hour. Adams, in the Handbook of Practical Program Evaluation, writes that "about one hour is considered a reasonable maximum length" for a semi-structured interview, to limit fatigue on both sides (Adams, 2015). Plan 45 minutes and let the good ones run to 60.

Build rapport fast, then get out of the way. DeJonckheere and Vaughn's advice is small and specific: dress professionally "but not overly formal", use "a normal conversational tone", and listen, "the key to successful interviewing". Your eyes belong on the participant, not on your laptop.

Let the order go, and follow the digression. If the answer to question two covers question five, skip ahead and say so. When a participant goes somewhere unplanned and it's rich, as Mariam's supervisor did, use the Follow probe and write "emergent: [topic]" in the margin. When it's not rich, use the Redirect probe. You decide in the moment, which is why the guide has to be short enough to hold in your head.

For more on the craft, our guide on how to conduct an interview was written for reporters, and almost all of it transfers. If your participants are product users, see our user research interview question bank.

Consent, recording and ethics approval

Most universities and many companies review a study before you interview anyone: an institutional review board, a research ethics committee or an internal privacy review. Find out which applies to you and build in the lead time, because the application usually wants your guide, your consent form and your plan for storing recordings.

Whatever the process, the interview itself runs on one habit: explain, ask, wait for a clear yes, and get the yes on the recording. Say who will hear the audio, how long you'll keep it and how the participant will appear in the write-up (a pseudonym, a role, a number). Recording law varies by place; our guide to one-party consent states covers the US state rules for a conversation you are part of.

Record, and record twice. DeJonckheere and Vaughn recommend audio recording so the interviewer can focus on the conversation and, when you interview alone, "two recording devices for each interview". A phone running ParrotNotes and a second recorder on the table is the usual setup; our guide on how to record an interview covers placement, Do Not Disturb and the on-tape opening script.

Want the transcript draft ready before you leave the building? Download ParrotNotes free: 100 minutes of recording a month, an AI summary on every recording, and nothing to carry but the phone you already own.

After the interview: memo, transcribe, code

Write the memo first. DeJonckheere and Vaughn put it plainly: "interviewers should always record memos", notes about what you are learning from the data. Before you check your email, dictate three minutes into ParrotNotes: what stood out, what surprised you, what to change in the guide, and any emergent theme you flagged. The dictation becomes a text note you can search later, which is more than most handwritten memos get.

Transcribe at the level your analysis needs. Thematic analysis needs every word and little else; conversation analysis needs timed pauses and overlaps. Decide once for the whole study and follow our guide to qualitative interview transcription for the protocol and the anonymization log. ParrotNotes gives you a transcript draft to correct against the audio rather than a blank page, which matters when you have 15 interviews and a supervisor.

Find the quote, then code. Ask the recording in ParrotNotes "Where did she describe the workaround for the night pick list?" and get the passage without scrubbing through 50 minutes of audio. On Pro, speaker identification labels the two voices, recordings run up to 3 hours, interviews in any of 99+ languages are transcribed and can be translated into the language you analyze in, and AI search finds "pick list" across every interview in the study. Then code: start with codes that match your core questions, and add the ones that emerged. Mariam's workaround started as a margin note, became a memo, then a code, then a chapter.

Build the guide, then let it go

A semi-structured interview in qualitative research works when the guide is short enough to leave behind. Write five to 10 core questions that each serve an aim, keep the probe bank in sight, pilot twice, stay under an hour, get consent on the recording, record twice, and write the memo first.

The method rewards listening, and listening is easier when you're not also the stenographer. Download ParrotNotes free before your next interview, press record with the participant's consent, and give the conversation your full attention. The transcript draft, the summary and the searchable memo will be waiting when you stand up.