UX Research · Step 3 · 6 min read
Notes, recordings and transcripts
An interview you can't find again might as well not have happened. How to take notes that keep what was said apart from what you think it means, when to record and how to ask, what a transcript is good for and where automatic ones go wrong, and the ten-minute debrief that saves your synthesis.
Your memory of an interview starts lying to you the moment it ends. By the next day, the participant's words have turned into your summary of them, and your summary is shaped by what you already believed. Notes, recordings and transcripts are how you keep the real thing. They're not admin: they're the raw material of every insight you'll write later.
What you'll be able to do after reading:
- Take notes in a grid that separates observation from interpretation
- Ask for consent to record, and handle recordings safely
- Use transcripts well, and check the automatic ones
- Run a ten-minute debrief and file everything so synthesis can start
1. Write what happened, then what you think
The most important habit in note-taking is keeping two things apart:
- Observation: what the participant said or did. Their words in quotation marks, as close as you can get. What they showed you, how long it took, where they paused.
- Interpretation: what you think it means. Useful, but it's yours, and it might be wrong.
Mix them and you can't go back. "Frustrated with the app" is an interpretation; "Sighed, and said 'here we go again' when the app opened" is an observation. Weeks later, in synthesis (see From notes to themes, affinity mapping), only the second one lets you check whether "frustrated" was right.
| Time | What they said or did | My interpretation | Tag |
|---|---|---|---|
| 12:40 | "I paste the bill into our group chat and everyone sends a thumbs up." Opens the chat to show it. | The thumbs up seems to mean "seen", not "paid". Check later. | RQ1 · tracking |
| 17:05 | Scrolls a shared spreadsheet for about 20 seconds to find March. Sighs. | The sheet is the real record, but hard to search. | RQ3 · workaround |
| 21:30 | "Honestly, I just pay it all and stop asking. It's not worth the fight." | Paying for peace: the cost here is social, not money. | RQ2 · conflict · quote |
| 29:15 | Says the landlord's water bill "comes whenever". Couldn't remember the last one. | Irregular bills may be the ones that get lost. | RQ2 · timing |
The grid makes the habit automatic. Time points back to the recording. What they said or did is the evidence. My interpretation is labelled as such. Tag links the note to a research question in your guide, so you can find every note on RQ2 later in seconds.
Tips for the note-taker:
- Write verbatim quotes in quotation marks, and only those. A paraphrase in quotation marks becomes a fake quote in a report.
- Write what people do, not just what they say. "Showed me the spreadsheet" is often worth more than five minutes of description.
- Mark surprises with a symbol you'll spot later. They're the notes you'll want first.
- Don't try to write everything. If you're recording, timestamps and keywords are enough to find the moment again.
2. Recording: ask, explain, protect
A recording lets you listen fully during the session and check your notes after it. It also creates personal data: a voice, often a face, and whatever the participant mentions about their life.
- Ask before you start, and explain. What you'll record (audio, screen, video), who will see it, how long you'll keep it, and that they can say no or ask you to stop at any time. Get the answer on the recording, too. NN/g's contextual-inquiry guide puts the same step in the first minutes of every session: discuss confidentiality and get approval for any filming or recording (NN/g, Flaherty, Contextual Inquiry, 2020).
- Record only what you need. If audio answers your research questions, don't record video.
- Store it where your organisation says, with access limited to the study team, and delete it when the retention period ends. Step 2 covers the rules: Research data and privacy and Research ethics and consent.
- Never share raw recordings widely. Share short, agreed clips or anonymised quotes instead.
If someone says no to recording, run the interview anyway. Your note-taker becomes essential.
3. Transcripts
A transcript turns an hour of audio into text you can search, quote and code. It's what makes careful Thematic analysis possible. Most video-call and research tools now transcribe automatically, which is a real gift, with three cautions:
- Check every quote against the audio before it goes into a report. Automatic transcription stumbles on names, accents, product terms and people talking over each other, and a wrong word can flip a meaning.
- Remove what identifies people: names, employers, places, anything that would let a reader recognise the participant. Label them P1, P2 and so on.
- Mind where the text goes. Pasting transcripts into an AI tool sends personal data to a third party. Check what your organisation allows first; AI in UX research covers how to do it carefully.
A transcript doesn't replace your notes. The transcript is what was said; your notes hold what you saw, what surprised you and what you thought at the time.
4. The ten-minute debrief
Straight after each session, before the next one or the next meeting, the interviewer and note-taker sit for ten minutes and write down:
- Three things that surprised us.
- The best quotes, with timestamps.
- What this person told us about each research question, in a line each.
- What to change in the guide before the next session.
Ten minutes now saves an hour later, because by the fourth interview the first one has blurred into the others. The debriefs are also the first draft of your synthesis: patterns start to show by the third one.
5. File it so you can find it
At the end of the day, every session should have one folder (or one row) with the same pieces: the consent record, notes, recording link, transcript, debrief, and the participant code. Use the same naming every time (for example 2026-11-18_P3_bills). It sounds dull. It's what lets a colleague, or you in six months, find the evidence behind an insight (see Research repositories).
6. Common mistakes
- Notes that mix what was said with what you concluded.
- Paraphrases written inside quotation marks.
- Recording without asking first, or without saying who will see it.
- Trusting an automatic transcript for quotes without listening back.
- Leaving names and employers in shared transcripts.
- No debrief, so the first interviews blur into the last ones.
7. Vocabulary
| Term | Meaning |
|---|---|
| Observation | What the participant said or did, recorded as faithfully as possible |
| Interpretation | What you think an observation means; kept separate and labelled |
| Verbatim | The participant's exact words |
| Transcript | A written version of the recording |
| Anonymisation | Removing details that would let someone identify the participant |
| Retention period | How long you're allowed to keep the data before deleting it |
| Debrief | A short meeting straight after a session to capture what stood out |
8. Try it (no tools needed)
- Watch five minutes of any interview video or podcast with a guest.
- Draw the grid on paper: time, what they said or did, my interpretation.
- Fill it in. Then cover the interpretation column and read only the observations: would someone else interpret them the same way?
Reflection questions: How many of your "observations" were really interpretations? Which quote did you write that you'd now need to check against the audio?
9. If you already practise
- Give note-takers a template with the grid and tags already set up for the study's research questions.
- Agree a set of tags before the first session, and add new ones only in the debrief.
- Keep a short clip library with consent for the clips, not the full recordings: three ten-second clips persuade a stakeholder more than a page of quotes.
- Audit last quarter's study folders: could you find the consent record for every recording?
Checklist
Learn more
- NN/g: Contextual Inquiry (Flaherty, 2020), for consent and recording at the start of a session · User Interviews 101 (Rosala and Pernice, 2023)
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.
Related
Running the session and listening · Research data and privacy · From notes to themes, affinity mapping · Thematic analysis · Research repositories · AI-assisted synthesis, step by step