Customer interview notes should preserve what the participant said and what the interviewer inferred as different kinds of information. Dictation cleanup must not turn a tentative interpretation into a direct customer claim.
Use voice capture to record fresh context, then review attribution and uncertainty before sharing. A clear note is valuable only if it remains faithful to the evidence.
Speak evidence and interpretation separately
Identify the participant by an appropriate non-identifying reference, the recent task, current workaround, reported frequency or spending, and the source of each statement. Label direct quotations only when wording is actually preserved.
Use explicit phrases for interpretation, such as interviewer hypothesis or follow-up needed. Avoid asking an editing system to make notes decisive when the conversation was uncertain.
Review a fictional scheduling interview
Suppose a participant says they sometimes lose track of requests and currently use a spreadsheet. The interviewer wonders whether automatic scheduling would help. The note should not become “the customer wants automatic scheduling” unless the participant actually said that.
Dictate separate sections for reported behaviour, a remembered paraphrase, observed friction, hypothesis, and next question. Check the cleaned version for changed frequency, removed qualification, invented willingness to pay, and false quotation marks.
Use harmless fictional material for the test. No customer research result or productivity measurement is claimed here.
Preserve context without unnecessary private data
Keep authorised source recordings or responses through the agreed retention and access process where they exist. A broadly shared product brief can use a de-identified summary with controlled references to evidence.
Review names, dates, numbers, consent to follow up, and any private details before export. A speech tool's local interface does not establish that audio or notes remain local, so understand the actual processing path.
The tradeoff is rapid capture versus evidence discipline. Speaking immediately after an interview can preserve useful context while memory is fresh, but it can also blend recollection and inference. A short structured review helps keep those categories visible.
Treat themes across interviews as analysis, not as automatic facts created by cleanup. Preserve disagreements and missing evidence rather than smoothing every respondent into one coherent narrative.
The StackFlight case study documents voice and text feedback collection with private review, while StackSpeech appears in the portfolio as a voice-input project. This scheduling interview is fictional and does not claim either tool automatically validates demand.
Questions about dictated interview notes
Can paraphrases be labelled as direct quotes?
No. Keep quotation status tied to actual preserved wording and use paraphrase labels when appropriate.
Should cleanup remove every uncertainty phrase?
No. Words such as sometimes, might, and not sure can be material evidence about the participant's meaning.
What should be checked before sharing?
Check attribution, quotations, numbers, frequency, hypotheses, follow-up permission, and private information. Bring ambiguous notes to a customer-research workflow discussion.