Advanced mode: train & predict separately

Use Advanced mode when the data you want to learn from and the sample you want to generate answers for are different files — e.g. train on last wave's full survey and predict the new wave's shorter questionnaire.

1. Upload the two files

In New Project, tick Advanced mode. Two dropzones appear:

• Training file — the complete dataset the models learn from.
• Predict file — the target sample whose missing questions will be filled with synthetic answers.

The columns the two files share are used to link them; the questions that exist only in the training file are the ones that get predicted.

2. Configure the project

Advanced mode opens a three-step flow (Load files → Configure project → Creating):

Field	Meaning
Project name	Optional — defaults to the file name
Folder	Every project lives in a folder
Multiple Choice Delimiter	Character separating multi-select answers (default _)
Age range / Country / Gender column	Map your demographic columns so the platform can use them
Questionnaire start	Where the questionnaire begins in the file

Click Next to create the project.

3. Start it manually

Unlike the simple flow, an Advanced project is created in CREATED state and waits for you — press the green play button in the project row when you are ready:

This gives you a chance to review the detected variables and adjust models per question (see Monitoring & results) before spending processing time.