Synthesize User Research with Claude
Turn raw interview notes, session recordings, and survey responses into prioritised findings in one session. Claude handles the pattern-matching; you handle the judgment.
The situation
You ran five user interviews. You have 40 pages of raw notes and a findings report due Friday. Manual affinity mapping takes hours you do not have.
You'll turn raw notes into a structured findings report with themes, participant counts, and verbatim quotes, then challenge it before you present anything.
What you walk away with
A thematic synthesis of your raw notes with participant counts and verbatim quotes per theme
Three to five prioritised design implications ready to take into a sprint
A challenge pass that flags sampling artifacts before you present to stakeholders
The difference one prompt makes
Don't
Summarise these user research notes.
Do this
Here are verbatim notes from 6 user interviews. Research question: why do users abandon the checkout flow? Identify the top 3 themes, include one verbatim quote per theme, and flag any theme raised by only one participant.
Prepare your raw notes for paste
Claude works best on structured raw material. Before pasting, add a one-line header to each participant's notes with their ID, role, and key context. You do not need to clean up the notes themselves: verbatim quotes and fragmented observations are fine.
# Format for paste-in synthesis
## P01 | Role: Small business owner | Device: Android, 4G | Session: 45 min
[raw notes verbatim here]
## P02 | Role: Freelance driver | Device: Feature phone, 2G | Session: 30 min
[raw notes verbatim here]
# Keep the headers consistent -- Claude uses them to
# cross-reference findings across participantsRun the synthesis prompt
Give Claude a specific synthesis task, not a general "summarise these notes" request. Specify the output structure you need: themes, supporting quotes, participant counts, and a severity or frequency marker. A structured output is easier to review and easier to present.
Here are my raw interview notes. Identify the top 5 themes.
For each theme: name it, write one sentence describing it,
list the 2-3 quotes that support it, and rate confidence
(high/medium/low based on how many participants mentioned it).
Do not invent themes that are not in the notes.Challenge the synthesis before presenting it
Claude finds patterns in what you gave it. It cannot know if your participant sample was skewed, if a quote was taken out of context, or if a theme is real versus a product of your recruitment criteria. Before you present the findings, interrogate them.
Play devil's advocate on theme 2. What evidence in the notes
contradicts it or suggests a different interpretation? What would
a skeptic say about this finding?Common mistakes
Four failure modes in research synthesis with Claude.
- Cleaning the notes before pasting them. Paraphrasing your own notes before synthesis removes the raw language participants used and replaces it with your interpretation. Paste verbatim. Messy is fine.
- Not specifying the research question. "Summarise these notes" produces a surface-level summary. "Identify themes related to why users abandon the checkout flow" produces findings you can act on. Always give Claude the research question before running the synthesis.
- Presenting the synthesis without the challenge step. Claude will find the most statistically prominent patterns in your notes. It cannot tell you which patterns are sample artifacts, which quotes are outliers, or which themes your recruitment criteria inflated. The challenge step is not optional if you are presenting to stakeholders.
- Using Claude to generate themes you then present as your own analysis. The synthesis is a starting point for your judgment, not the final analysis. If a theme Claude found does not ring true to you based on being in the room, flag it as tentative or drop it. Your presence in the sessions is data that Claude does not have.
What's next?
Automate repetitive design tasksNew guides, when they ship
One email, roughly weekly. CLAUDE.md templates, workflows I actually use, and the cut-for-length stuff that does not make the public guides. One-click unsubscribe.
Or follow on Substack