Choose a starting point

QubeSurvey is built for people who are comfortable writing a questionnaire as a script in an editor. That is a different starting point from a drag-and-drop form builder: there is no browser builder yet, and the first step below needs Visual Studio Code.

If that is not you, read Do I need a code editor? first. It is short, and it will save you an afternoon.

Choose a path below: test an existing questionnaire, build more involved survey logic, or draft questions for a new study. Each starts with local testing. When the questionnaire is ready, follow these four steps:

  1. Write and test locally. Install the QubeSurvey extension and use the qube command line with an AI agent you choose. Check with invented answers.
  2. Upload to app.qubesurvey.com for team testing and tickets.
  3. Run fieldwork. Use an open link or personal invitations when enabled.
  4. Export and download the data. Use an available Run export.

The first-survey guide explains each step and the collection settings.

Before you start

You need Visual Studio Code 1.102 or later and permission to install an extension. Install the QubeSurvey extension. Local writing and interview testing need no QubeSurvey account. Use invented answers while testing. If you want help from an AI agent, follow Use qube with an AI agent.

Do I need a code editor?

Yes, today. A questionnaire is a .qube text file. You write it in VS Code with the QubeSurvey extension, and that is also where you run a local interview test. There is no browser form builder.

Three honest cases:

  • You already use VS Code, or you are willing to install it. Start with the first-survey guide. An AI agent that can run qube can draft most of the script for you.
  • Someone technical can set it up for you. Ask them to install the extension and create the first Script, then you can edit the wording and run interviews.
  • You want a drag-and-drop builder with no install. QubeSurvey is not that product yet. Google Forms, Microsoft Forms or Typeform will get you a survey today; come back when you need versioned review, controlled fieldwork or an export you can defend.

I have a straightforward survey and do not use a code editor

Start with the first-survey guide, but expect to install VS Code first. If you can describe your questionnaire in plain words, an AI agent that can run qube can write most of the script; you review the questions and test the routes. The free parts — writing, local testing, team preview and manual tickets — do not need a hosted Account or a payment.

I work in market research or script surveys for NIPO

Start with a study specification and its existing questionnaire.

  1. Keep the specification beside the script. Choose a part to check, such as screening for eligibility or asking follow-up questions about selected brands.
  2. Open the existing script. For QubeScript, see the quickstart. For an existing Odin file such as p1234q, save an editor copy as p1234q.odin for VS Code integration and test the behavior you need. See Using existing Odin scripts.
  3. Open the file in the QubeSurvey extension. Resolve errors and review warnings.
  4. Answer through that part of the survey. Compare the wording, answer options and next questions with the specification. Use Back to change an earlier answer and check what happens next.

Keep a note of what you tested and what still needs checking before using the whole questionnaire in fieldwork.

I need more control over survey routing

Start with survey logic that involves several questions. For example, ask which transport people use, then offer only those answers when asking for their main mode of travel. If they go back and remove an option, the follow-up choices should change too. The commuter-travel example shows this in a working questionnaire. For a larger study with loops and calculations, choose an example study.

  1. Write and test the survey logic: Describe the questions, which answers affect later questions, and what should happen when an answer changes. Adapt an example in VS Code, with your AI agent and qube if useful. Try several answer combinations and use Back to check corrections.
  2. Upload for team review: Have teammates test that version in their browsers.
  3. Record a problem: File a ticket against the tested version. Describe the expected and actual behavior and relevant conditions.
  4. Review a proposed fix: A teammate or your own AI agent can upload a corrected version and mark the ticket fixed. Compare the source and test the affected and nearby routes, including corrections with Back, before using the version. See Testing a script for saved journeys and replay.

Keep the tested script and the answer combinations you used so the team can repeat the checks after a change.

I am exploring a basic research question for a startup

Start with a decision you need to make and the people you need to hear from. For example, if you are considering a delivery service, ask potential customers how they currently receive deliveries and what problems they encounter.

  1. Write down the decision, the target group, and what you still need to learn.
  2. Use the tiny QubeScript sample as a prototype. Change only what is needed to ask your first questions.
  3. Run the prototype locally in QubeSurvey extension with invented answers.
  4. Review the wording for leading questions. Check that the target group is appropriate and decide how you would recruit them; the prototype does not establish either.

If you use an assistant you already have, under your own model and data settings, you can give it this brief:

I am deciding [decision] for [product or service]. The people whose views matter are [target group]. I need to learn [one or two unknowns]. Suggest three to five neutral draft questions for a short prototype, based on the QubeScript sample or study I provide. Mark assumptions and flag wording that could lead an answer. Do not claim the questions are validated or choose who to recruit. Return suggested wording only; I will review it and write or adapt the QubeScript source.

Treat generated wording as a draft, not proof of research validity. Review the wording, target group and recruitment plan yourself.

You now have draft questions to review and test before planning recruitment.

Continue with a Run

When the local questionnaire is ready, follow Getting started to upload it for team testing, create a Run and download a test response.

Uploading requires sign-in and permission to edit the Script. The available sign-in methods depend on the site; email sign-in links are limited to existing accounts and current project invitees. See People, access and billing for creating an Account and sharing Script access. The hosted qube commands need qube login; see hosted commands.