Creating from scratch

Create a survey from scratch when a ready-made template isn't suitable: you need your own topic, your own wording, or a localized diagnostic for a specific situation. For your first regular launch, it's often better to use a template: questions and analysis topics are already assembled there.

Launch workflow

Open the Surveys section inside the organization and click Create survey. In the scenario selection window, select Create survey manually, enter a name, and click Go to editor.

The editor will open right after the draft is created. First, click Add category. A category groups questions of one block: for example, "Team", "Processes", or "Leadership". Inside the category, click Add question or Generate question if you want an AI draft.

In the question editor, fill in the wording, check the answer type, answer options, intent, and instructions for AI. Click Save question — without this, changes to the question will not be applied.

When the structure is ready, open the survey card and click Launch survey. After confirmation, the public link will become active, and you can send it to participants.

What to consider

Before launching, a manual survey remains a draft: participants do not see it and cannot respond. If the survey card says The survey is currently empty, return to the editor and add questions.

A minimum of 3 questions is required to launch. The upper limit is 36 questions; in practice, it's better to keep the survey shorter so that participants reach the end and leave detailed comments.

By default, the audience is the entire organization. If you need a different group of participants, configure it before launching and check the rules in the Survey audience article: the audience is fixed upon creation and cannot be changed for an already launched survey.

After the first responses

In a manual survey, analysis topics appear during finalization: the AI examines real responses and constructs a coherent picture from them. That's why clear question wording and understandable instructions for the AI are especially important.

When enough responses are collected, run the finalization. You can perform an intermediate analysis without closing the measurement, or close the measurement and start the final processing.