A copyable, method-agnostic template for keeping an audit trail in qualitative research — the dated record of analytic decisions that lets a supervisor, examiner, or reviewer trace your path from raw data to findings.
Audit trails are the practical backbone of the dependability and confirmability criteria described by Lincoln & Guba (1985). In a viva, the difference between "trust me" and "here is exactly how I got there" is this document.
One table, five columns, kept from day one:
| Date | Stage | Decision | Rationale | Scope / data affected |
|---|---|---|---|---|
| 12 Mar | Coding | What you decided | Why — in one or two sentences | Which data/codes it touches |
Copy audit-trail-template.md into your research journal, a spreadsheet, or your analysis software. 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.
- 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 (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
ai-decision-log-template.md
A worked reflexive thematic analysis example with six realistic entries is in examples/. A full study might have 40–80 entries.
Audit trails come from Lincoln & Guba's trustworthiness framework, while Braun & Clarke are wary of imported "reliability" logic in reflexive TA. The resolution is in the framing: in RTA, this 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 & Clarke ask you to demonstrate. Same table, different epistemological job.
"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. The ai-decision-log-template.md covers that layer.
New in v1.1: codebook-templates/ — seven CSV codebook and coding-scheme templates shaped by the logic of their methodology (codebook TA, reflexive TA theme log, Gioia data structure, grounded theory coding, content analysis coding scheme, IPA experiential statements, template analysis), each with realistic filled example rows and a near-miss column where the method calls for one. The codebook records what your coding scheme is; the audit trail records how and why it changed.
- Full step-by-step guide with worked example
- Supervisor trust pack — a plain-language pack for documenting AI involvement for your committee
- Audit trail explainer (trustworthiness criteria)
Maintained by QualIntel OS — AI-assisted qualitative analysis software built around the principle that every interpretive decision stays with the researcher, recorded in an exportable audit trail. This template is free to use and adapt under CC BY 4.0; the discipline it encodes doesn't require any particular software.