Interview notes are the hardest kind, because the person is watching you take them. Every second you spend writing is a second of broken eye contact, and the follow-up question you did not ask because you were still writing down the last answer is the one that would have produced the useful part.
This applies whether you are hiring, running user research, or interviewing a source. The structure below is built around one principle: capture as little as possible during, and reconstruct immediately after.
Prepare the page before they arrive
The most effective thing you can do happens before the interview starts. Write your questions down the left side of a page, with space under each.
This does two things. Your notes now have a skeleton, so during the conversation you are filling gaps rather than composing structure. And you stop losing your place, which is the usual reason interviewers ask a question that was already answered.
For hiring, structure is not just convenience. Structured interviews, where every candidate gets the same questions and is rated against defined criteria, are consistently better predictors of job performance than unstructured conversation, and they are far more defensible if a hiring decision is ever challenged. The prepared page is what makes the structure real.
Leave a column on the right for reactions of your own. Keep them visibly separate from what was said, using brackets or a different indent. Confusing your inference with their words is the single most damaging error in interview notes, and it happens most often when writing up from memory later.
During: capture five things, not everything
You cannot transcribe and interview at the same time. Pick a small target and let the rest go.
| Capture | Why |
|---|---|
| Exact quotes, in quotation marks | The one thing memory cannot rebuild. Two or three per interview is plenty |
| Numbers and names | Precise, unrecoverable, easy to get wrong later |
| Contradictions | They said X, then implied not-X. Mark it, come back at the end |
| Their words for things | The vocabulary they use is data, especially in user research |
| Your follow-up prompts | One word in the margin so you can return without interrupting |
Everything else is summary, and summary can wait fifteen minutes. What cannot wait is the question you thought of at minute six.
The margin prompt deserves emphasis. Interrupting to chase a thread kills the thread they are on. Writing → why the spreadsheet? in the margin lets you finish listening and still ask.
Recording, and the honest version of asking
Record if you can, and ask plainly. "Do you mind if I record this so I can focus on listening rather than writing?" is an easy yes for most people, and the reason given is true.
Two rules once it is running. Ask before you start recording, not after, in every jurisdiction and regardless of local law, because the trust cost of doing otherwise is not worth it. And still take notes. A recording you have to re-listen to in full has not saved you time, it has moved the cost. Timestamps and quotes on the page turn 45 minutes of audio into three lookups, the same trick as taking notes from a video.
If you are the only interviewer and the conversation is high-stakes, a second person taking notes is worth more than any technique on this page. One interviews, one writes.
The ten minutes afterwards decide everything
This is where interview notes are actually made. Do not schedule interviews back to back, and if you must, leave ten minutes between them.
Immediately after, before you speak to anyone:
- Fill in the gaps. Your fragments are still decodable now, and will not be tomorrow.
- Write the three-line summary. What did this person actually tell you? If you cannot say it in three lines, you have data and no finding.
- Separate what they said from what you concluded. Explicitly. Two headings if necessary.
- Log your immediate impression, dated. In hiring, your view will drift as you meet other candidates and as you talk to colleagues. The uncontaminated version is worth having on record.
The urgency is not a stylistic preference. Memory for conversation decays fast and it decays by reconstructing, filling gaps with what you expected to hear rather than what was said. Ten minutes now beats an hour tomorrow, for the reasons set out in the forgetting curve.
Writing up for other people
Interview notes usually have a second audience, which changes what a good note looks like.
For user research, resist summarizing into conclusions too early. The value is in specifics: what they were trying to do, what they actually did, what they said in their own words. Themes come from reading across many interviews, and a note pre-compressed into "user wants better search" has destroyed the evidence. Keep the raw quotes.
For hiring, write to the criteria you defined beforehand, evidence first. "Rated 3 on system design: walked through a sharding trade-off unprompted, could not explain failure modes when asked" is useful. "Seemed strong technically" is not, and will not survive scrutiny.
For both, be careful what you record about a person. Interview notes are often discoverable, and in many places candidates can request them. Write things you would be comfortable having read back to you.
The pattern across many interviews
One interview is a note. Twenty interviews are a body of evidence, and only if they can sit beside each other. Use the same headings every time, tag them consistently, and give each one a title that says what came out of it rather than just the person's name.
Finding the pattern across them is a synthesis problem, not a note-taking problem. Maps of content and concept mapping both help, and the underlying discipline is in how to take research notes.
More situational guides in how to take notes. For the meeting case, which has different rules, see how to take meeting notes and how to take 1:1 notes.