Customer interview synthesis

From a pile of transcripts to insights you can defend

Your AI synthesizes interview notes the way a user researcher would: observations atomized and tagged, grouped by meaning, counted by participant, contradictions surfaced, and insights with quotes, confidence levels and next steps.

By Vander Labs. A Vander Labs Original, written by the Vander Labs team with AI assistance. Works with Claude, ChatGPT, Cursor and Claude Code.

What your AI learns

• Atomizes transcripts into tagged observations with participant codes
• Groups by meaning and counts distinct participants, never mentions
• Writes insights with evidence, confidence and what would raise it

How it works

1. Inventory the data: Participants, segments, dates and recruiting bias, so every claim is framed.
2. Atomize: One fact, behavior or quote per line with a participant code; no interpretation yet.
3. Tag: Topic and type tags, with workarounds and behaviors weighted above stated desires.
4. Group by affinity: Clusters named in the participants' language, merged or split until each holds one idea.
5. Count and weigh: Distinct participants and segments per cluster; single-segment patterns flagged.
6. Contradictions: Disagreements, behavior versus attitude, and the expected things that never came up.
7. Write insights: Cluster plus interpretation plus implication, with two quotes, a count and a confidence level.
8. Answer and recommend: Each research question answered or not, then next steps sized to the decision.
9. Keep the trail: An observations CSV so every insight traces back to the words that produced it.

Limits and safety

• Reports counts of participants, not percentages of a market, and states sampling bias in every confidence level.
• Observations and interpretation are kept in separate sections.

Topics: user-research, interviews, synthesis, affinity-mapping, insights