Methodological rigour
How to Write an Audit Trail for Qualitative Research
Published 4 July 2026 · Includes a copyable template
An audit trail is a dated record of the analytic decisions you made and why — the document that lets a supervisor or examiner trace your path from raw data to findings. It is the practical backbone of the dependability and confirmability criteria set out by Lincoln and Guba, and in a viva it is the difference between "trust me" and "here is exactly how I got there."
This guide covers what to record, a template you can copy today, a worked reflexive thematic analysis example, and what changes when AI tools are involved.
What an audit trail records
- Design decisions — methodology chosen and rejected alternatives, sampling logic, question refinements — with reasons.
- Coding decisions — every code created, renamed, merged, split, or retired, and the reasoning. This is the layer examiners probe hardest.
- Theme-level decisions — how codes were grouped, how candidate themes evolved, what was collapsed or discarded between theme-map versions.
- Reflexive notes — where your positionality plausibly shaped an interpretive choice, and what you did about it.
- AI involvement (if any) — what the tool suggested, what you accepted or rejected, and on what grounds — see below.
The template
One table, five columns, kept from day one. Copy it into your research journal, a spreadsheet, or your analysis software:
| Date | Stage | Decision | Rationale | Scope / data affected |
|---|---|---|---|---|
| 12 Mar | Coding | What you decided | Why — in one or two sentences | Which data/codes it touches |
The discipline is in the rationale column. "Merged two codes" is a log line; "merged because both describe material obstacles and the separation was fragmenting the pattern" is an audit trail.
Worked example — reflexive thematic analysis
| Date | Stage | Decision | Rationale | Scope / data affected |
|---|---|---|---|---|
| 02 Mar | Design | Chose reflexive TA over codebook TA | RQ asks how participants make sense of service barriers — interpretive depth over coder agreement | Whole study |
| 09 Mar | Coding | Created code participant_voice | Recurring pattern: participants describing not being consulted; semantically distinct from access barriers | T01–T04 |
| 14 Mar | Coding | Merged access_cost + access_transport → barrier_structural | Both describe material obstacles, not attitudes; separation was fragmenting the pattern | 11 segments, 6 transcripts |
| 21 Mar | Coding | Rejected AI-suggested segment for coping_peer | Excerpt describes family support, not peer support — conceptually distinct in this dataset | T07, lines 214–221 |
| 28 Mar | Themes | Collapsed candidate themes 'system failure' + 'system coordination' into one | On review against coded extracts, both described the same organising concept from different vantage points | Theme map v3 |
| 04 Apr | Reflexivity | Noted my practitioner background may foreground service-critique readings | Re-read T02/T05 checking for confirmation bias; retained interpretation with widened extract base | Themes 1–2 |
Six entries covering design, coding, themes, reflexivity, and one AI rejection. A full study might have 40–80 of these. That volume sounds like work — until you are three weeks from submission trying to reconstruct why a theme changed in April.
A note on Braun & Clarke
Researchers searching for a "Braun and Clarke audit trail" sometimes hit an apparent contradiction: audit trails come from Lincoln and Guba's trustworthiness framework, while Braun and Clarke are wary of imported "reliability" logic in reflexive TA. The resolution is in the framing: in RTA, the record is not evidence that your coding was "correct" — it is a transparent account of how your interpretation developed, which is exactly the researcher-owned subjectivity Braun and Clarke ask you to demonstrate. Same table, different epistemological job.
If AI touched your analysis, the bar is higher
"I used AI-assisted tools" satisfies no examiner in 2026. If an AI surfaced candidate evidence or suggested codes, your audit trail needs a layer a manual study doesn't: which suggestions you accepted, which you rejected, and why. That log is what separates AI-assisted analysis you can defend from AI-generated output you have to hope nobody questions — and it pairs with a proper AI disclosure statement in your methods chapter.
For what this looks like as trustworthiness criteria, see the audit trail explainer; for method-specific expectations, the thematic analysis guide.
Or let the trail write itself
QualIntel OS builds this table automatically: every code change and every accepted or rejected AI suggestion is timestamped and attributed as you work, then exported as an examiner-ready evidence pack with the disclosure statement generated from the log — not written after the fact.
Start freeFrequently asked questions
What is an audit trail in qualitative research?
An audit trail is a documented record of the decisions made throughout a qualitative study — from research design through coding to theme development — with the rationale for each. It lets a supervisor, examiner, or reviewer trace how you moved from raw data to findings, supporting the dependability and confirmability criteria described by Lincoln and Guba (1985).
What should an audit trail include?
At minimum: dated entries recording (1) design decisions and their rationale, (2) coding decisions — codes created, merged, renamed, or retired and why, (3) theme-level decisions — how codes were grouped and how candidate themes changed, (4) reflexive notes on how your positionality shaped interpretation, and (5) if AI tools were used, a log of what the tool suggested and what you accepted or rejected.
Do I need an audit trail for reflexive thematic analysis?
Braun and Clarke position reflexive TA within a qualitative paradigm where 'accuracy'-style measures like inter-coder reliability don't fit — but they consistently call for researchers to document and own their analytic decisions. In RTA the audit trail works best framed as a reflexive decision record: not proof that coding was 'correct,' but a transparent account of how your interpretation developed.
What is an example of an audit trail entry?
A dated entry might read: '14 Mar — Merged codes access_cost and access_transport into barrier_structural. Both describe material obstacles rather than attitudes; keeping them separate was fragmenting the pattern. Affects 11 coded segments across 6 transcripts.' Short, dated, decision plus rationale plus scope.
How is an audit trail different from a reflexive journal?
They overlap but serve different functions. A reflexive journal records your evolving thinking, assumptions, and reactions to the data. An audit trail records concrete analytic decisions and their justification. Many researchers keep one document with both layers; the audit trail is the part an examiner can follow step by step.