Qualitative analysis
This address used to describe an AI-assisted coding task, a budget, and qualitative exports. The product does not do that work. What it does is collect comments and, after collection closes, let two assigned people code them. An adjudicator settles a comment only where those two disagree.
The procedure is Coding and adjudication. This page says what that task is, and what is absent, so the old outline is not read as the current one.
Comments the study collects
A rating can include a comment when Setup asks for one: never, when the rating is low, or when the rating is low or the person cannot rate. Comments are not a separate analysis product. They are part of the answer.
On a later round, panelists see comment text only when the team chose to share it. Sharing follows the study's recorded choice: word for word without names, word for word after identifying details are removed, edited, summarized into themes, or not shown. That choice is not an AI summary.
Human coding
Coding waits until collection has closed. A save while the round is still collecting is refused. A finished round no longer accepts coding changes.
On Study team the slots are Coder 1 (human), Coder 2 (human), and Settles disagreements. The two coders must be different people. The person who settles disagreements cannot also be a coder. Coders work independently and do not see each other's codes. Panelists appear by number only.
The comments page is headed Comments on a rating round and Coder comments on a questionnaire round. A coder ticks the ideas in each comment and uses Save codes. Where the codes differ, the assigned adjudicator uses Settle it, gives a reason of at least 10 characters, and uses Settle the disagreement.
People allowed to see agreement get Agreement between the coders, including Kappa. Kappa is reported. It does not stop a round, change a result, or decide a statement. When the study has no minimum agreement set, nothing is checked against one.
The product does not run an AI qualitative-analysis pipeline. There is no budget for model use. No suggestion is generated. A code that is saved was chosen by a person in a human coder seat.
Analysis plan can store Allow AI to help with comments, off by default, and the labels under it, including Suggest codes for comments, as Coder 3. Study team can show the heading Coder 3 (AI). The comments page does not offer those suggestions. Saving the preference is not a coding step, and it is not approval of generated text. Whether a later version should run a model is an open product question. This page does not describe that work as a task you can do.
Let researchers export raw comments on Analysis plan stores a preference. It does not create a comment file.
Exports this task does not produce
The product does not build a qualitative ZIP, a REFI-QDA file, a PDF, a DOCX file, or a checksummed bundle of comments.
On Reporting, Codebook is listed as "CSV · Current version". Export stays unavailable. The reason is "No codebook has been saved for this study." The response file from Reports says whether a comment was given. It does not contain the comment text.
The files that do exist are on Reporting and exports.
Next step
Assign the two coder slots and the adjudicator on Study team, then follow Coding and adjudication after you close collection.