Prepare variable names before collection.
Use stable, understandable codes and keep them distinct from the wording shown to participants. A short code identifies the variable; a full label explains the question.
In Nuvopoll, AI can suggest missing variable names when the template is saved. You can enter your own. The naming step uses question information, not participant answers.
The later SPSS export is a separate server operation.
Take a labelled file and its supporting documents.
| Output | Check before analysis |
|---|---|
| SPSS .sav | Variable labels, value labels, data types and defined missing codes. |
| CSV response table | The coding and missing-value interpretation in the software you open it with. |
| Codebook | Which question and options each variable represents. |
| Flow diagram | Which participants were eligible to see conditional questions. |
The CSV is usable in Excel but is not a native .xlsx workbook. Keep the survey version, language and recruitment source when they are relevant to the analysis.
Missing values are not all the same.
- −99: not shown because of the question path.
- −98: shown but unanswered.
- −97: participant declined to answer.
The .sav defines these as missing values. Confirm how your statistical procedure uses them and do not include them as numeric observations. A conditional support question applies to support users, not the full customer sample.
Check multiple responses and open text.
Inspect how each selected option is represented in the export before defining a multiple-response set or calculating a percentage. Decide whether the denominator is people or total selections.
Read text answers and transcription outputs as text. Inspect identifying details before sharing a working dataset. A word cloud or exported transcript is not a completed qualitative analysis.
Make a small test export first.
Enter test responses that cover different branches, refusal, partial completion and language versions. Open the export and codebook together. Check labels, options and missing codes before real collection.
The raw-response exports support further statistical work. Measurement-model results currently stay in Nuvopoll and are not exported as a model-result file; their scope and limits are described on the analysis page.