Treat the transcript as data, not admin

By the time an interview has finished, the recording feels like the precious thing. In practice, the transcript is what your analysis will live in for months. Supervisors, reviewers and ethics committees will read it, quote from it and occasionally challenge it. That means the strength of your argument often rests on decisions made by whoever typed the words, which is why it pays to treat transcription as part of the method rather than a task to be cleared. A transcript that preserves meaning, marks its speakers clearly and follows an agreed style guide will serve you far better than one that is merely neat.

Agree your style guide before anyone starts typing

Most transcription problems are not typing problems; they are agreement problems. Before the first file is opened, settle how the transcript will look. A short written style guide of one or two pages saves hours of correction later and makes comparison across interviews possible.

• Verbatim level. Decide whether you want true verbatim (ums, false starts, repetitions), intelligent verbatim (lightly tidied for readability) or a clean version intended for publication.
• Speaker labels. Choose a convention — for example P1 and P2 for participants, INT for the interviewer — and keep it identical across the whole project.
• Timestamps. State whether you need one every minute, at each new topic, or only alongside extracts you plan to publish.
• Layout. Consistent spacing, a plain font and clear indentation make long documents far easier to code and quote from.

Send the guide to anyone involved in transcription and ask them to flag anything it does not cover. Small ambiguities, such as whether to record laughter or a long pause, are much easier to resolve once than repeatedly.

Identifying speakers and untangling overlap

Clear attribution is the single most useful thing a transcript can offer. If a reader cannot tell who said what, a quotation becomes unusable, however accurate the wording. Label speakers from the very first line and keep those labels stable; switching from "Participant A" to "P1" halfway through creates needless doubt about who is speaking.

Where two people talk over one another, resist the temptation to guess. Mark the overlap with a consistent device — square brackets are common — and indicate where one speaker continues underneath another. For group discussions, add a short roster at the top listing who was present, in what role, and roughly when they spoke. Readers then arrive at the quotes already oriented, and you avoid a paragraph of explanation before every extract.

Filler words, dialect and the limits of tidying

Removing every "um" and "you know" can make a participant sound more fluent than they were, which is a small act of interpretation. Many style guides permit light tidying for readability, but only where meaning is untouched. The safe rule is this: tidy hesitations, never content.

• Do not correct grammar or complete a sentence the speaker left unfinished.
• Do not standardise dialect, regional vocabulary or pronunciation spellings — they may carry meaning you will later want to discuss.
• Do keep markers such as laughter, sighs and significant pauses, because they often do more work than the words around them.
• When quoting for publication, signal omissions with ellipses and any editorial additions in square brackets.

If a speaker uses a term in a way that looks like an error but may be deliberate, leave it and raise a query note rather than silently "fixing" it.

Anonymisation, ethics and sensitive passages

Consent forms usually promise anonymity, and the transcript is where that promise is most easily broken. Replace names with agreed pseudonyms or codes, and store the key separately and securely. Watch for indirect identifiers — job titles, small towns, distinctive family circumstances, named colleagues — which can identify someone as surely as a surname.

Decide in advance how to handle distress, disclosures, or anything said after the recorder was switched off. Where speech is genuinely inaudible, write [inaudible] rather than inventing plausible words, and where a passage is withheld, say so plainly rather than leaving a silent gap. A clear audit trail matters if an ethics committee later asks how the material was managed.

Checking the transcript before it leaves your desk

Always spot-check against the audio. Read a random five minutes while listening; you will catch dropped words, misheard numbers and swapped speakers that a silent proofread would miss. Then work through the following before you file the document.

• Listen back to the opening and closing minutes, where names, consent statements and any closing remarks usually sit.
• Search for placeholder text — [inaudible], [??], [check name] — and resolve every instance.
• Confirm speaker labels match your roster and that no line has been attributed to the wrong person.
• Check that any quotation you intend to publish reads identically to the version in the transcript you are citing.

If the transcript will be deposited in an archive or appended to a thesis, confirm it meets the required standard before submission. Finally, keep the audio. A transcript is a reading of a recording, not a replacement for it, and a last read-through aloud will reveal where the sense breaks down in a way no spellcheck ever will. Your future self, deep in coding and quoting, will be grateful for the care.