Coding your team can do together.
Qualitative analysis is collaborative work trapped in single-seat desktop software. CORPUS puts the transcripts, the codebook and the coders in one place: agreement calculated as you go, AI first-pass coding you can overrule, every code traceable to the person or model that applied it — and none of it on anybody's laptop.
Two researchers code the same twenty interviews, in two copies of the same desktop program, on two laptops. Then somebody has to merge the files. That is where qualitative rigour usually dies.
The established tools are good at coding and structurally bad at teams: a licence per seat, a project file per person, a merge step everyone dreads, and sensitive transcripts sitting on hardware nobody has assessed.
CORPUS starts from the opposite assumption: coding is a group activity with disagreement in it, and that disagreement is data about your codebook, worth seeing while the coding is still under way.
It also reads transcripts straight from TRANSCRIPTION, so the audio never leaves your perimeter on the way in either.
Where the disagreement shows up.
An excerpt from a real interview shape: a patient describing why they stopped taking a medication. Watch the codebook build as passages are coded, then watch what happens at the passage the two coders read differently. That flag is deliberate.
Interviewer Can you tell me about the point where you stopped taking it?
P14 Well, the tiredness was the thing. I would get home from work and just sit down, and that was the evening gone. I have two children and I could not do that to them for another year.
Interviewer Did you talk to anybody about it?
P14 I rang the department twice and left a message, and nobody rang back for a fortnight. By then I had decided, and honestly I did not want to be talked out of it.
Interviewer And how do you feel about that decision now?
P14 It was my choice and I would make it again. Though if somebody had told me it might settle after three months, I might have waited.
Four weeks from fieldwork to a codebook you trust.
Material in
Audio transcribed inside your perimeter, or existing transcripts imported with their speaker labels intact.
- TRANSCRIPTION, or your own files
- Case attributes from your database
- Pseudonyms applied at import
Codebook v1
Built from your protocol or grown from the first interviews, with definitions written before coding starts.
- Deductive, inductive or both
- Definitions with examples
- Agreed by the whole team
Double-code and calibrate
A sampled subset coded independently by two people. Agreement is read as evidence about the codebook, and the codebook is fixed.
- Kappa per code as well as overall
- Adjudication with recorded outcomes
- Definitions revised, versioned
Code the corpus
The rest of the material, with the AI first pass carrying the recognisable passages and humans deciding all of them.
- Progress and saturation tracked
- Memos as you go
- Provenance recorded per code
Themes and writing
Themes named by researchers, supported by retrievable quotations, with the reporting checklist filled from what actually happened.
- Quotations with stable IDs
- COREQ / SRQR generated
- Audit trail for the reviewer