How QubeSurvey works
QubeSurvey is built on a few ideas. Knowing them explains why the product behaves as it does, and what it will not do.
A questionnaire is a text file
A survey is a script: a readable text file holding the questions, the answer options and the routing. You keep it where you keep other work, compare versions of it, and let an AI agent read and edit it with you. There is no second copy of the survey inside a form builder.
When you upload a script it becomes a Version. A Version never changes. To correct a questionnaire you upload a new Version; the old one stays exactly as it was tested.
A Run uses one Version
A Run is a round of data collection. It uses the Version you chose and keeps using it. Uploading a newer Version does not change a Run that is collecting, so a response always means what the tested questionnaire said it means.
The script decides what is asked and what an answer means. The Run decides how the study is fielded: who can take it, when it is open, and which endings keep their answers. See Managing fieldwork.
An interview is a record of what happened
What a respondent was shown, what they answered and how the interview ended are recorded as they happen. Response lists, counts and export files are read from that record. They can be produced again at any time, and they cannot disagree with it.
Two consequences:
- An answer that was not given is recorded as not given. It is never stored as zero or as an empty text.
- The outcome settings in force when an interview started are the ones that apply to it. Changing a Run’s settings later affects interviews that start afterwards.
Nothing is claimed that cannot be shown
- A script is checked before it runs. Syntax the system would ignore is rejected with a diagnostic instead of being accepted and doing nothing.
- A respondent whose page cannot be submitted is told why, and their answers stay on the page.
- Before a Run goes live, a readiness list states what would stop it and where to fix it.
- The guides name only commands and settings that exist; a test compares them with the code.
If you find a place where QubeSurvey says something happened and it did not, or stays silent when something failed, that is a defect. Please report it.
You pay for data you can take away
- Writing and testing a script on your own computer is free and needs no account.
- An accepted response is charged when its answers are kept. An ending that keeps nothing is free and returns no data.
- Every kept response appears in the export, and exporting is free.
- Credit is bought in advance. There is no subscription and nothing renews by itself.
See People, access and billing for the rates.
It is meant to stay small
The product has a short list of things to understand: Account, Script, Version, Run, Participant, Group, Service. A feature is added when it replaces something or when a real study needs it. That is also why some things other survey tools offer are absent, and why Choose a starting point says so plainly.
Where to go next
- Getting started: write, test, run and export one survey.
- Use qube with an AI agent: the command line an agent uses to check and test a script.