Separate user behaviors from user opinions in interviews

Sort raw user interview transcripts into verifiable past actions and unverified personal opinions to prevent product bias.

Prompt

Prompt
Act as a rigorous user research analyst. Analyze interview notes to separate verifiable past actions from subjective opinions, hypothetical promises, and polite compliments. Rules: Classify every user statement as either Observed Past Behavior or Stated User Opinion; explain the product design risk of treating each opinion as a concrete user requirement; do not assume future behavior based solely on an opinion. Output format: 1. Behavioral Breakdown Table (Columns: Quote or Note, Classification, Evidence Level, Product Risk if Misinterpreted), 2. Confirmed Usability Insights (bullet list of actual behaviors), 3. Research Gaps to Validate (bullet list of questions to confirm opinions with data). The interview notes are below:

Example output

Example for: transcript from an interview on an invoice management tool

1. Behavioral Breakdown Table
Quote or NoteClassificationEvidence LevelProduct Risk if Misinterpreted
"I export everything to Excel every Friday to calculate totals."Observed Past BehaviorHigh (describes recurring current habit)Low; this confirms existing workflow friction and manual workarounds.
"I would easily pay an extra $30/month if you added automated tax calculations."Stated User OpinionLow (hypothetical future statement)High; building a paid tax feature based on stated intent often yields near-zero adoption.
"Your interface looks much cleaner than the old platform we used."Stated User OpinionLow (polite compliment)Medium; aesthetic praise may mask serious underlying workflow bottlenecks.
2. Confirmed Usability Insights
  • Users actively export CSVs weekly because native summary reporting does not aggregate taxes across multiple legal entities.
  • Workflow completion requires switching between our software and external spreadsheet tools.
3. Research Gaps to Validate
  • What specific calculation formulas are users executing in Excel that native reports fail to calculate?
  • Does export volume correlate with billing tier, or is it widespread across all customer sizes?
Written by AI from this prompt. Your results will vary.

Paste this prompt into Claude, ChatGPT, or Gemini, then paste your raw user interview text at the bottom. Follow up by asking the model to suggest two analytics events to track in production to verify the stated opinions.

Make it repeatable

Keep using this prompt without pasting it in every time.

Keep exploring

More in Meetings & Notes