Creating a project¶
The standard flow: one survey file in, one augmented dataset out.
1. Upload your survey file¶
Go to New Project and drag your Excel survey file into the dropzone (or Click to select).
The file should be a survey export with one row per respondent and one column per question. See Preparing your data file for the coding conventions that keep the system from having to ask you questions before the project can start.
2. Pre-processing (optional)¶
If your data needs work before modeling, tick any of the pre-processing steps:
| Option | What it does |
|---|---|
| Clean & reshape | Fixes rows, columns and headers |
| Analyze numeric fields | Snaps or drops numeric values by divisor |
| Normalize open ends | Normalizes open-ended answers: brands, concepts, sentiment |
| Creative Insights | Decomposes images into attributes |
| Advanced mode | Upload train & predict files separately — see Advanced mode |
Each option is covered in depth in Pre-processing.
3. Pick or create a folder¶
Every project must belong to a folder — the basic organizational unit of your account. Old projects may still appear on the general Projects screen, but new projects are always created inside a folder. Choose an existing one or create a new one with the folder button:
4. Model mode¶
On the Configure project step, next to the folder picker, Model mode decides which models fill the synthetic answers:
| Option | What it does |
|---|---|
| Classic (Random Forest ensemble) | The default. Leave it selected unless someone on your team tells you otherwise. |
| New (KNN calibrated) | Copies answers from the closest real respondents with exact quotas — faster, closer marginals and crosstabs, and no near-copies. |
The choice is made once, at creation, and applies to the whole project. Per-field choices in Model tuning still override it afterwards — see Prediction engine.
Not sure which to pick?
Keep Classic and use Automatic mode (next step) instead: it tries several versions and keeps the best one for you. The Model mode selector may be removed in a future release in favour of Automatic mode.
5. Automatic mode (optional)¶
At the end of the wizard, tick Automatic mode if you would rather not choose settings yourself. n-Infinite then:
- generates five versions of the synthetic data, each from a different random seed (the number that fixes the run's random choices);
- scores each version internally and delivers only the one with the best quality score;
- applies the standard delivery checks and the logical consistency repair automatically — you do not need to turn them on.
It takes longer than a standard run, because the data is generated several times. The result is judged like any other project: see the Quality certificate on the project detail screen.
The Start processing button enables once the file and folder are set.
6. Start processing¶
Click Start processing. The project is created and training starts automatically — you are taken to the Projects list where you can follow its progress. See Monitoring & results.