User Interviews: A Private Workflow from Recording to Searchable Quotes

User interviews: A private workflow from recording to searchable quotes

Three weeks after an onboarding study, you remember a participant explaining why they stopped halfway through setup. You need that evidence for a product discussion, but you can’t remember which interview it came from or whether your notes captured their words or your interpretation.

Reusable evidence needs a source you can return to and enough context to distinguish what the participant said from what you inferred. This workflow follows one fictional onboarding study from recording to retrieval, keeping audio on your device.

What user interviews reveal, and what they cannot prove

A user interview is a one-on-one research conversation about someone’s experiences, needs and motivations. Use it to explore an unfamiliar workflow or understand a reported problem before deciding what to investigate or build.

Interviews let you probe unexpected answers. Surveys collect responses at scale, with less room for follow-up. Usability testing observes someone attempting tasks; an interview captures their account of the experience. Hearing someone describe setup does not demonstrate that they can complete it. Hearing a problem repeatedly does not establish how common it is across your user base.

Formats range from structured interviews with fixed questions to unstructured conversations guided by a broad topic. For this study, use a semi-structured guide: prepare core questions, then allow follow-ups. Start with “Tell me about the last time you set up your account.” Follow their sequence rather than asking them to endorse a proposed fix.

Set up the study so evidence stays searchable

Give the fictional study one question: “What causes new account owners to pause setup before their first data import?” Assign participants P01, P02 and P03. Use those IDs in filenames and notes instead of names or email addresses.

Create a study folder on your device:

2026-10_onboarding-first-import/
 01_raw-audio/P01.m4a
 02_transcripts/P01.txt
 03_evidence-notes/P01.md
 04_synthesis/findings.md

Repeat the participant files for P02 and P03. Keep the ID-to-identity mapping outside this collection, with restricted access. Pseudonyms reduce casual exposure; recordings and conversation details may still identify someone.

Use consistent descriptive tags: stage:data-import for passages about the import step and task:connect-source for connecting a data source. These labels help you find material without remembering the participant’s wording.

Keep interpretations such as unclear-data-permissions in a separate theme field in synthesis. A passage about connecting a source does not automatically support that explanation. Themes remain claims you can revise.

Record with permission, then preserve the source

Before recording, explain what you’ll capture, who can access it and when it will be deleted. Replace the brackets in this introduction with your actual arrangements:

“We’re studying where account setup becomes difficult. With your permission, I’d like to record this conversation so I can check your wording later. The recording and transcript will stay on this computer, accessible to [researcher or named roles]. We’ll delete them by [date]. You can decline recording or ask me to stop. Are you comfortable with that?”

Wait for permission before starting. MeetingDebrief captures microphone and system audio locally, without uploading the recording or adding a bot to the call. For remote interviews, test that both sides are audible. In person, check microphone placement with the participant seated where they’ll speak.

After recording, check playback and preserve an unedited source copy before correcting transcript errors. Apply the retention deadline to that copy and any backups.

On-device processing does not replace permission or device security. Use encryption and a screen lock, restrict file access, and check whether automatic cloud backup includes your study folder.

Turn each transcript into a reusable note

Store multiple excerpt records in each participant file, such as 03_evidence-notes/P01.md, rather than compressing the interview into one conclusion. Include these fields in every record alongside the fictional template below:

Tags: stage:data-import, task:connect-source
Quote verification: unchecked transcript

Study: 2026-10_onboarding-first-import
Participant: P01
Research question: What causes new account owners to
 pause setup before their first data import?
Contextual observations: Participant recalled stopping
 at the source-connection step; no task was observed.
Exact quote: “I could connect it, but I wasn’t sure
 whether that would pull in all our customer records.
 I waited for our admin to check.”
Source: 01_raw-audio/P01.m4a, 12:40-12:53
Researcher interpretation: Uncertainty about import scope
 may have prompted an approval check.

A hypothetical automated summary might say “Participant couldn’t connect their data source.” That turns uncertainty about consequences into inability to act. Use summaries to locate passages, not as quotations. Mark the quote verified only after listening to the source.

In MeetingDebrief, each recording becomes a session containing audio, transcript, summaries and notes. Summaries are generated on-device from the transcript. Per-meeting notes are searchable and included as chat context, so label interpretations explicitly: later answers may draw on them alongside the transcript.

Reuse these fields across studies. When copying excerpts into synthesis, retain the study identifier and source path so October’s P01 cannot be mistaken for a later participant.

Retrieve a quote, then check its context

Weeks later, the team is considering a connection-error fix. You remember someone waiting for approval rather than reporting an error.

Open MeetingDebrief’s global search with Cmd+K on macOS or Ctrl+K on Windows. Try is:meeting admin, then is:meeting customer records. Search covers transcripts, summaries, notes and dictations; is:meeting excludes dictations. If you’ve tagged the sessions, add tag: followed by the study tag. Tags on external evidence files do not filter app sessions.

During study setup, put a study-and-participant identifier such as 2026-10_onboarding-first-import P01 in each session’s searchable per-meeting notes. Copy that identifier into the external evidence note as Session reference, and record the preserved source path, 2026-10_onboarding-first-import/01_raw-audio/P01.m4a, in both places. Keep session-level study tags separate from excerpt-level labels such as task:connect-source. For retrieval, search the identifier and check the source path before selecting Chat with this meeting. Keep the request within that interview:

Locate the passage where the participant discusses waiting for an admin before connecting a data source. What concern did they describe? Cite the timestamp and distinguish transcript wording from researcher notes.

Answers include timestamp citations. Clicking one jumps the audio and transcript to that moment. Before quoting:

• Listen for exact wording, including qualifications or negation the transcript may have missed.
• Confirm the voice is the participant’s, not the interviewer paraphrasing. Check the participant ID against your study record.
• Play the preceding question and following exchange. Check whether the concern arose spontaneously, followed a leading prompt, or was later corrected.

Build a finding from several interviews

This fictional evidence matrix adds two contrasting accounts. Timestamps are illustrative.

Study: 2026-10_onboarding-first-import. All quotes below have Quote verification: unchecked transcript; check each against its source audio before use.

A provisional theme is that uncertainty about import scope can delay connection for some account owners. Test a clearer pre-connection explanation of which records will be imported rather than committing to a connection-error fix. Include people whose source data is not ready, so the test can distinguish clarity from readiness. Report two supporting accounts and one contrasting account in this study, not a percentage of customers.

For a Markdown handoff, MeetingDebrief can export transcripts and summaries. Check that the output suits your workflow. Manually carry the study ID, source filename and timestamp into the finding rather than assuming exported links preserve them.

Before handing over a finding, check:

• Can a reviewer open every supporting and conflicting source?
• Does the evidence-note template include Quote verification: unchecked transcript? Change it to checked against source audio only after listening and correcting the quoted wording.
• Does 04_synthesis/findings.md include an Unresolved disagreements field recording competing interpretations and their sources, or none recorded?
• Does the proposed test address the theme without ignoring alternative causes?

Try the workflow with one consented interview. A week later, find a passage without opening your evidence note first. Record what slowed you down, such as a missing search term or inaccessible source, and fix it before the next study. MeetingDebrief’s 7-day trial needs no account or card.