A study's states

A study goes through three states, and each one allows something different. Almost every time the product will not let you touch something, the answer is here.

Draft

Every study is born here, and in this state everything can change: the name, the dates, the scope, the model. The wizard resumes where you left it, and a draft can be re-edited in full through all four steps, not just its name.

A draft asks nobody anything. It does not exist for employees.

Launching

The transition from draft to in progress. Two things are required: the sample must already be generated, and your plan must allow it (see Account and plan).

On launch, three things freeze until the study closes:

• The scope. Which node is measured and, with it, everything below.
• The instrument. Each branch's questionnaire and its questions. You also cannot reassign the questionnaire of a branch this study is measuring.
• The confidentiality threshold. The one in force is locked in. Changing it later does not touch this study.

At that same moment the period's targets are set. See Improvement targets.

The freeze is not administrative rigidity: it is what makes the result mean something. If the questionnaire changes halfway, the first half of your people answered one thing and the second half another, and both groups get published together as if they were one.

Practical corollary: change anything you want to change while it is a draft. After launch, the only way out is waiting for it to close.

In progress

The study is collecting. Here is the part that surprises people most:

A study in progress shows NO results. Only progress.

They are not hidden by the threshold: they are not computed. Results are produced on finalization, and until then what exists is participation.

And that is deliberate. Looking at partial results of a climate survey invites conclusions about whoever answered first, and those are never a sample of anyone — they are the most motivated, or whoever was at the tablet at nine in the morning.

What you do see

Collection progress: how many people have responded, broken down by node, plus the pace and an estimate of when it closes.

Progress is not subject to the confidentiality threshold. The threshold protects survey results, not progress: knowing that three out of four people in a store responded says nothing about what they answered.

Each person's state

«Unavailable» is what keeps collection from staying open forever. Someone who has left will never respond; if they kept counting as pending, the study would never reach the end. Marking them removes them from the expected without inventing an answer they never gave.

Finished

On finalization, results are computed and stored. From then on you have both indices per branch, the breakdown by category, segments, and —from the second study onward— the historical series.

Results are persisted, not recomputed every time you open the screen. That is what makes a result published last year still say the same thing today, even if you have since reorganized the tree or moved people between nodes.

Summary

If the product will not let you do something

Before hunting for a permission, check the state. The two usual causes are that the study has already launched, or that your role does not write to that resource — and they are easy to tell apart: if it is the state, the button is there but fails or warns; if it is the role, the button does not exist. See Roles and permissions.