AIWorkDex
Customer Research Practice

Build traceable customer research findings

Preserve customer evidence, test early themes, and connect each finding to a decision or next question.

Preparation
An analysis round after de-identification
Difficulty
Advanced
Task
Turn interviews into findings
Back to the Task: Turn interviews into findings

Before you start

Get what you need before you start

  • De-identify approved transcripts and notes before adding them to analysis tools.
  • Keep protected originals available for authorised source checking.

Steps

Work through the method

Use the listed inputs and tools. Check the evidence when you need to verify a step.

  1. Preserve the source trail

    Assign a participant code and stable locator to each transcript, note, or observation record.

    Why it matters: Every later interpretation needs a route back to the protected source.

    Input
    Approved research records.
    Output
    A source inventory with stable locators.
    Tools in this stepGoogle Sheets

    Evidence for this step

  2. Extract one observation at a time

    Record what a participant said or did, its locator, and the context without adding interpretation.

    Why it matters: Observations should remain separate from the finding they may support.

    Input
    The source inventory and protected records.
    Output
    Atomic, source-linked observation rows.
    Tools in this stepChatGPT data analysisGoogle Sheets

    Evidence for this step

  3. Verify every extracted row

    Compare each AI-assisted or manual extraction with the protected source and correct wording, context, and locator.

    Why it matters: File analysis can structure material, but its output still needs human source checking.

    Input
    Extracted observation rows and protected sources.
    Output
    Human-verified observation rows.
    Tools in this stepChatGPT data analysisGoogle Sheets

    Evidence for this step

  4. Group related observations

    Place related rows together while keeping participant codes and source locators visible.

    Why it matters: Grouping should preserve the evidence behind a possible theme.

    Input
    Verified observation rows.
    Output
    Evidence groups with provisional labels.
    Tools in this stepGoogle Sheets

    Evidence for this step

  5. Test provisional themes

    Search the full set for exceptions, competing explanations, and segments where the pattern does not hold.

    Why it matters: Early themes can hide contrary evidence.

    Input
    The provisional groups and full observation set.
    Output
    Theme notes with supporting and contrary rows.
    Tools in this stepGoogle Sheets

    Evidence for this step

  6. Write bounded findings

    State the observed pattern, context, supporting evidence, contrary evidence, and inference limit.

    Why it matters: A finding needs a clear boundary between evidence and interpretation.

    Input
    The tested theme notes.
    Output
    Traceable finding statements.
    Tools in this stepGoogle Sheets

    Evidence for this step

  7. Choose the action or next question

    For each finding, record the decision it informs or the evidence gap that needs another round.

    Why it matters: Analysis ends when a finding informs work or clarifies what remains unknown.

    Input
    The traceable findings and pending decision.
    Output
    A decision record or next research question.
    Tools in this stepGoogle Sheets

    Evidence for this step

Success checks

Check the result before you finish

  • Every observation has a stable source locator.
  • Every AI-assisted extraction has a recorded human check.
  • Every finding includes contrary evidence or states that none appeared in the reviewed sample.

Failure modes

Watch for these problems

  • Remove unapproved personal or sensitive data before using any analysis tool.
  • Return to the protected source when a row loses context.
  • Keep a theme provisional when a segment or contrary case could change it.

Tools

Choose the tools you need

Sources

Read the sources behind this practice

Check what each source supports and where the advice has limits.