Quick answer
- Gather all notes and transcripts. Write the research questions where everyone can see them.
- Write one observation or quote per note, labeled with the participant code. Sort the notes into groups and name each group.
- Count people, not mentions: "4 of 8 participants" is evidence; one person saying it six times is not a pattern.
- Write each insight as observation + evidence + implication, and check each quote against the audio.
Before you start: gather the data and the team
Do synthesis soon after each round of sessions, with the people who watched them. The GOV.UK Service Manual says to invite anyone who observed the research and to analyse "while it is still fresh in people’s minds." It gives a planning rule: "Aim to spend 1 hour analysing every 2 hours of research." For eight one-hour interviews, use four hours as an initial analysis budget and allow more time if needed.
- Collect one file per participant. Notes, transcript and recording, all named with the same participant code. A notes layout is in User Interview Notes Template (Free, Copy-Ready).
- Write the research questions on the wall. Nielsen Norman Group suggests this before the team reads the data, so the questions are easy to refer to while you work.
- Read everything once. Highlight what looks important, but do not group yet.
Affinity mapping: sort observations into groups
Nielsen Norman Group defines affinity diagramming as "organizing related observations, ideas, concepts, or findings into distinct clusters." On a wall or a digital whiteboard:
- Write one observation per note. The Service Manual asks people to "write exactly what they saw or heard (eg verbatim quotes or observed behaviour), not what they think it means." Add the participant code and the time to each note.
- Sort the notes into groups. Move notes that are about the same thing next to each other. Anyone can move anyone's note.
- Name each group. Use a short label, such as "checks totals outside the app". Split a large group into smaller themes.
- Keep small groups. Nielsen Norman Group says: "Don’t discount a small cluster just because it has few sticky notes." A single strong observation can point to a risk.
How to analyze user interviews: code each passage
Affinity mapping works from notes. Coding works through the full transcripts, so you are less likely to miss a passage. Nielsen Norman Group describes a code as "a word or phrase that acts as a label for a segment of text." Some codes describe the topic, such as "invoice matching". Others add your reading, such as "distrust of totals".
Keep a short codebook: the code name, one line on what it means, and one example quote. In academic work, a widely cited approach is Braun and Clarke's reflexive thematic analysis. Their site describes thematic analysis as "a range of qualitative research methods that are used for exploring and interpreting patterned meaning across datasets." Transcripts for coding are covered in How to Transcribe Interviews for Qualitative Research.
Count across participants, not mentions
Nielsen Norman Group says a theme "emerges when related findings appear multiple times across participants or data sources." So count people:
- Write counts as "n of N". "5 of 8 participants checked totals in a spreadsheet." Avoid percentages for small groups.
- Count each person once per theme. One talkative participant can fill a whole cluster.
- Note who did and who said. "3 of 8 did it in the session; 2 more said they do it."
- Record the counter-evidence. "2 of 8 trusted the totals" belongs next to the theme, not in the bin.
A count shows how widespread a pattern is in your sample. It does not show how serious the pattern is, or how common it is in your whole user base.
Write insight statements: observation, evidence, implication
The Service Manual asks for findings "as short statements that summarise what you’ve learned." One of its examples is "people think they can click the progress bar to navigate". Add two more parts:
- State the observation. What people did or said, in plain words.
- Add the evidence. The count and two or three quotes with participant codes and times.
- Add the implication. What it means for the product or the next round of research.
Observation: Office managers re-check invoice totals outside the app before they pay.
Evidence: 5 of 8 participants opened a spreadsheet first; P03 at 00:04:10: "I check this first, every time."
Implication: the payment screen does not yet earn trust; test showing the matched invoice lines before the pay button.
User interview synthesis template (worksheet)
Copy one table per theme.
| Field | What to write |
|---|---|
| Research question | The question this theme answers. |
| Theme | Short group name from the affinity map. |
| Observation | What people did or said, without opinion. |
| Participants | Codes and count, such as P01, P03, P05, P06 (4 of 8). |
| Quotes | Two or three, each with participant code and time, checked against the audio. |
| Counter-evidence | Who did not fit, with codes. |
| Insight statement | Observation + evidence + implication, in one or two sentences. |
| Next step | Design change, new research question or backlog item, with an owner. |
How searching across transcripts helps
Nielsen Norman Group warns against analysis that focuses "on only memorable events and quotes." Search lets you test a theme against every session. Before you finish a worksheet, search all transcripts for the theme's key words, such as "spreadsheet", "double-check" and "paper". You find participants you did not count and quotes that go against the theme.
The built-in route: Voice Memos can transcribe speech on iPhone 12 or later in supported languages and regions. Apple says its search list "shows only recordings with that text—either in the title or in the audio transcription." Select a word in the transcript and the playhead jumps to that place in the audio. A search hit, or an AI answer, tells you where to look. It is not evidence until you listen to the passage.
How to summarize interview findings for the team
Keep the summary to one page: the research questions, the top three to five insight statements with counts, one quote for each, the counter-evidence, and the next steps. The Service Manual lists actions such as new design ideas, new research questions and new user stories for the backlog.
HOW LOCUS FITS
Ask one question across every session
Locus records on Apple Watch or iPhone, and the iPhone app transfers the recording to your Mac.
The Mac app transcribes it, writes a summary, can remember the voices of people you choose, and answers questions across your recordings, such as "Who checked totals in a spreadsheet?" All of that runs on the Mac with local AI; no cloud AI is required. Device transfer can use an encrypted relay, with Remote Access enabled by default. A cloud-synced notes folder still syncs. An external AI service is an optional, separate choice; when it is on, that text leaves the device.
An answer is a pointer: you still count the participants yourself and check every quote against the audio. The app is in preview. It needs a Mac with Apple Silicon M3 or later, 24 GB of unified memory or more, and macOS 27 or later. Watch recordings need the iPhone app.
Join the waitlistQuestions people ask
How do you synthesize user interview data?
Gather all notes and transcripts. Put one observation or quote on each note with the participant code. Sort the notes into named groups, code the transcripts for anything the notes missed, and count how many participants support each theme. Write each insight as observation, evidence and implication, then check the quotes against the audio.
What is the difference between affinity mapping and thematic analysis?
Affinity mapping sorts individual notes into clusters, on a wall or a digital board. Thematic analysis tags passages in full transcripts with codes, then builds themes from the codes. It takes longer, but you are less likely to miss a passage. You can combine them: an affinity map first, then coding to check and fill in the themes.
How do you summarize interview results?
Use one page. List the research questions, then three to five insight statements. Give each one a count, such as "4 of 8 participants", one checked quote with a participant code, and the implication. Add the counter-evidence and the next steps with owners. Link every insight to its worksheet so readers can trace it back to the evidence.
What makes a good insight statement?
It has three parts. The observation says what people did or said. The evidence gives the count and two or three quotes with participant codes and times. The implication says what the team should change, test or ask next. A statement without evidence is an opinion; a statement without an implication is a fact that nobody acts on.
Can Locus help with user interview synthesis?
It can help you find evidence. Locus records sessions on Apple Watch or iPhone, transfers them to your Mac, and transcribes and summarizes each one with local AI. It answers questions across your recordings, so you can ask which sessions mention a topic. You still sort the notes, count participants and check each quote against the audio. The app is in preview.
Sources
- Analyse a research session (GOV.UK Service Manual) — team analysis, one observation per note, affinity sorting, findings as short statements, checked 2026‑10‑23.
- Affinity Diagramming for Collaboratively Sorting UX Findings and Design Ideas (Nielsen Norman Group) — definition, steps, small clusters, checked 2026‑10‑23.
- Analyze Qualitative Data: Thematic Analysis (Nielsen Norman Group, Maria Rosala) — codes, themes across participants, research questions on the wall, checked 2026‑10‑23.
- Thematic Analysis (Virginia Braun, University of Auckland, and Victoria Clarke, University of the West of England) — definition of thematic analysis, reflexive TA, checked 2026‑10‑23.
- View a Voice Memos transcription on iPhone (Apple Support) — transcripts on iPhone 12 or later, search in transcriptions, checked 2026‑10‑23.