Marcus Andrey Study
Ler em português Library

UX Research · Step 3 · 7 min read

The interview guide

A guide is not a script: it's a short plan that keeps a free conversation on the questions your study has to answer. How to build one from your research questions, order it as a funnel, give it a warm-up and a wrap-up, and pilot it before the first real session.

Two chairs face each other across a small table with a recorder on it, and each chair has a speech bubble above it. Point at the participant's chair and their bubble opens wide, line after line; point at the interviewer's and theirs only ever grows to a short question. That's this step in one gesture: in an interview, the other person does the talking and you listen. This step has five lessons: the interview guide, asking good questions, running the session and listening, notes, recordings and transcripts, and contextual inquiry and diary studies.Illustration: Marcus Andrey, drawn with Hairline (MIT, Lucas Marques).

A good interview feels like a conversation, and that's exactly why it needs a plan. Without one, you follow the most interesting story and come back with nothing on the question your team needed answered. With a script, you read questions in order and never hear the story at all. The interview guide sits in between: a short page that tells you what you must cover, and leaves you free in how you get there.

What you'll be able to do after reading:

  • Explain the difference between an interview guide and a script
  • Turn research questions into open interview questions
  • Order a guide from warm-up to core to wrap-up, as a funnel
  • Pilot a guide and fix it before the first real session

1. A guide, not a script

In Short user interviews you met the basic idea: five or six open questions as a safety net. This lesson builds the real thing.

Most UX interviews are semi-structured: there's a plan, but the interviewer can follow up on anything important the participant says. NN/g's distinction is the useful one: unlike a script, an interview guide can be used flexibly. You can ask questions in any order, skip some, and ask ones that aren't written down (NN/g, Rosala, Writing an Effective Guide for a UX Interview, 2021).

So why write it at all? Because without one, you risk not covering the topics your research questions need in every interview, and then you can't compare across participants. NN/g says plainly that it's not unusual to spend a full day on a good guide. It's worth it: it's the day that makes the next ten hours useful.


2. Start from the research questions

The guide is downstream of the plan. Your research questions are what the team needs to learn (see Research questions that lead to decisions); your interview questions are what you actually say to a person. They're never the same sentence. "Where does bill-splitting go wrong?" is a research question. Asked like that, a participant will give you an opinion. Asked as "Tell me about a time splitting a bill didn't go as you hoped", you get a story.

NN/g's process for building a guide, in short (Rosala, 2021):

  1. Write down the research questions.
  2. Brainstorm every interview question that comes to mind, good or bad.
  3. Make them broader. Most first drafts are closed questions. For each one, ask: is there a broader, open version? "What time do you start work? Who do you work with?" can become "Describe a typical day at work", which will probably cover both.
  4. Check each question against a research question. If a research question has no interview question, write one. If an interview question serves no research question, cut it.
  5. Order them, from easy to reflective, often in the order the experience happened.
  6. Add follow-ups under each question, so you remember to dig.
  7. Pilot the guide.

Interview guide: sharing household bills

Study: how flatmates and couples split bills todayLength: 45 minResearch questions: RQ1 how they decide who pays · RQ2 where it goes wrong · RQ3 what they use to keep track

Introduction0–5 min

  • Who I am, what the study is for, how long it takes. "There are no right answers: we're learning from your experience."
  • Consent: recording on? Notes anonymised, kept until the study ends. Questions before we start?

Warm-up5–10 min

  1. Tell me a little about where you live and who you live with.
  2. What does a normal month of bills look like in your home?

Core10–38 min

  1. Tell me about the last time a bill arrived and had to be split. RQ1What happened next? Who decided? How did you let the others know?
  2. Tell me about a time splitting a bill didn't go as you hoped. RQ2What did you do about it? How did that feel? How did it end?
  3. Walk me through how you keep track of who owes what. Can you show me? RQ3When did you start doing it that way? What do you do when it doesn't work?
  4. If not covered: How often do you check it? Who else looks at it? RQ3

Wrap-up38–45 min

  1. Is there anything about bills at home I should have asked, and didn't?
  2. Thank you. What happens next, and how the incentive is paid.
A filled-in guide for a 45-minute interview. Each core question opens with a real past event, has its follow-ups underneath, and is tagged with the research question it answers, so a gap shows before the first session. Closed questions wait at the end, only if the story didn't cover them.Example study and guide: Marcus Andrey, CC BY 4.0. Structure after NN/g (Rosala, 2021).

Notice the small tags next to each core question. That's step 4 made visible. When you read the guide the night before, a research question with no tag is a gap you can still fix.

See Interview guide for the same structure on one page, with timings and a template to copy.


3. The shape: warm-up, core, wrap-up

Introduction (about 5 minutes). Say who you are, what the study is for and how the information will be used, then ask for consent to record (Research ethics and consent). NN/g advises slowing down your speech here: talking slowly has a calming effect and shows you have time to listen (NN/g, Rosala and Pernice, User Interviews 101, 2023).

Warm-up (about 5 minutes). Easy, open questions that get people talking: "Tell me a little about yourself", "What does a normal month look like?". Avoid anything that might feel personal or judgemental. NN/g's example: "What was the last book you read?" assumes they read one recently, and if they didn't, they might feel stupid (User Interviews 101).

Core (most of the time). Three to five broad questions, each one about a specific past event, with its follow-ups underneath. This is where your research questions get answered.

Wrap-up (about 5 minutes). One open door ("Is there anything I should have asked?"), then thanks, what happens next and how the incentive is paid. The open door question regularly brings out the thing that mattered most to them.


4. Order it as a funnel

  1. Broad, open storyTell me about the last time a bill had to be split.
  2. Open follow-upWhat happened next? Why did you do it that way?
  3. Narrow, openHow did you let the others know what they owed?
  4. Closed, for a detailWas that the same app as last month?
The funnel: every topic starts broad and ends narrow. Broad questions let people bring up what matters to them before you've shown what you're interested in; closed questions only fill in details. The funnel starts again with each new topic.Diagram: Marcus Andrey, CC BY 4.0, after NN/g, The Funnel Technique (Rosala and Moran, 2022).

Inside the core, each topic goes from broad to narrow. NN/g calls this the funnel technique: start with broad, open questions, follow up with open probes, and only then use closed questions to fill in details. The reason is bias: ask specific questions too early and you show what you're interested in, and people start answering that instead of telling you what happened (NN/g, Rosala and Moran, The Funnel Technique, 2022).

In practice: if the participant covers your closed questions in their story, you never ask them. That's why they sit at the bottom of each topic, marked "if not covered".


5. Pilot it

Even the best guide needs changes after the first interview. NN/g recommends a pilot: try the guide with a colleague or friend if the topic isn't specialised, or with one or two real target users if it is (User Interviews 101). You're looking for questions that confuse, questions that get one-word answers, and a guide that doesn't fit the time.

Small changes during the study are fine; interviews are qualitative. What you shouldn't change mid-study are the research questions, or you won't have enough data on any of them by the end.


6. Common mistakes

  • Writing a script and reading it in order, like a survey said out loud.
  • Copying the research questions into the guide as interview questions.
  • A core section full of closed questions ("Do you…?", "How often…?").
  • Asking for opinions about a product before hearing how people live with the problem.
  • No time plan, so the most important topic gets the last five rushed minutes.
  • Skipping the pilot and debugging the guide on your first real participant.

7. Vocabulary

Term Meaning
Interview guide The plan for a semi-structured interview: topics and open questions, used flexibly (also called a discussion guide)
Script A fixed list of questions read in order, the same for everyone
Semi-structured interview An interview with a plan, where the interviewer can follow up on what matters
Research question What the team needs to learn; never asked to the participant as written
Funnel technique Ordering each topic from broad and open to narrow and closed
Warm-up Easy opening questions that build comfort and get people talking
Pilot A trial run of the guide before the real sessions

8. Try it (no tools needed)

  1. Take one research question you care about at work, or this one: "How do people decide what to cook on weeknights?"
  2. Write five interview questions for it. Then rewrite each closed one as an open one about a specific time.
  3. Put them in funnel order, add two follow-ups under each, and say the guide out loud with a timer. Did it fit in 30 minutes?

Reflection questions: Which of your first five questions were really research questions in disguise? Which one would you cut if you only had 20 minutes?


9. If you already practise

  • Put the research questions at the top of the guide and tag every core question with one. Coverage gaps show before the session, not in synthesis.
  • Keep a team library of warm-ups and wrap-ups that work; write new cores for each study.
  • When a stakeholder sends you "questions to ask users", translate them into research questions first, then write the interview questions yourself.
  • Review guides the way you review designs: one colleague, ten minutes, before the pilot.

Checklist

Learn more

Books, courses and videos to go deeper

Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.

Short user interviews · Research questions that lead to decisions · Writing a research plan · Asking good questions · Run three interviews · AI in UX research · Interview guide

Saved only on this device.

↑ ↓ to move · Enter to open · Esc to close