A respondent says they occasionally forget a follow-up, and the research note becomes “the team cannot manage enquiries reliably.” The note has strengthened both the frequency and the scope of the original claim.
For founders analysing voice feedback, preserve meaning before improving readability. A concise note should remain traceable to what the person actually said and the question they were answering.
What makes a research note faithful?
A faithful note separates a direct statement, a paraphrase and an interpretation. It retains important qualifiers such as sometimes, in one location or during a particular event.
Transcription and summarisation can each introduce errors. Review consequential details against the source rather than assume a fluent transcript or summary is accurate.
Use a source-to-note review table
The following fictional examples show how meaning can drift. They are illustrative responses, not findings from a real customer study.
| Source meaning | Faithful note | Overstated note |
|---|---|---|
| One missed follow-up last week | Respondent described one recent miss | Follow-ups are routinely lost |
| Unsure who buys software | Purchasing role remains unknown | Respondent controls the budget |
| Might try a reminder | Expressed tentative interest | Committed to using the product |
| Spreadsheet works for most tasks | Current workaround is partly effective | Existing workflow is broken |
Keep a source reference and relevant timestamp where appropriate. The reviewer should be able to inspect the original context without copying the whole recording into every document.
Preserve uncertainty and disagreement
If a phrase is unclear, mark it for review instead of replacing it with the most convenient interpretation. If two respondents describe different workflows, retain both rather than average them into a single fictional customer.
Do not convert tone into a claim about urgency or purchasing intent without supporting evidence. A confident speaker can still be describing a rare problem or a task outside their decision authority.
Make edits reviewable
Store the original response separately from the working analysis, with appropriate access. When a note changes materially, retain enough context to explain why the interpretation changed.
My StackFlight case study describes preserving original voice or text responses for founder review. It does not claim automatic analysis establishes the correct product decision.
The feature-request evidence guide helps keep interest, behaviour and commitment distinct during synthesis. Use that distinction when the summary will influence an engineering budget.
Finish by asking another reviewer to trace a proposed finding back to its source. If the supporting language is weaker than the finding, narrow the claim or collect additional evidence rather than improve the wording alone.
Should every note be a verbatim transcript?
No. Concise paraphrases are useful when labelled and faithful to the original meaning.
Can an AI summary replace source review?
Use it as an aid, with consequential interpretations checked against the original response and context.
What should we preserve first?
Keep the authorised source, question version and reviewable note before preparing a founder research brief.