The org tree

Censuma does not assume your organization has fixed levels. There is no «branch» field and no «department» field: there is a tree of nodes that you build, with the depth and the names your company actually has.

A retail chain might go Country → Region → Store → Department. A forty-person consultancy might have two levels and stop there. Both are valid structures, and the product treats them the same way.

One single root, and the account is not part of the tree

Each organization has exactly one root node, and you create it yourself with whatever name and type you want. It is not automatic and it is not the account: the organization is who we bill, and it never appears in the structure.

The distinction looks like a detail and it is not. The account is a data boundary — nothing from one organization is visible from another. The tree is an analysis tool: you change it when the company changes, or when you want to look at the results a different way.

The node is the unit of everything else

Almost everything in the product hangs off a node:

• Employees. Every person in the roster lives in a node.
• Questionnaires. A questionnaire is assigned to a node and its descendants inherit it, so you do not have to assign it store by store.
• The scope of a study. A study measures a node and everything below it.
• Permissions. A manager is assigned a branch and sees that branch: not the rest of the tree, and not the results next door.
• Results. Every index is published per node. That is what lets you say «the Northern Region is eight points below average».

The shape of the tree decides what you can compare

This is the part worth thinking about up front rather than afterwards.

Results are published per node. A cut that does not exist in the tree does not exist in the results: if all the cashiers in a store hang off the same node as the stockers, there is no way to look at them separately, however interesting the question. And the other way round — a tree split into twenty nodes of six people each produces twenty results, almost none of which can be published, because they do not reach the confidentiality threshold.

It is a trade-off between granularity and publishability, and it has a hard floor. Every node whose results you want to read has to gather enough responses to clear your organization's threshold. Before subdividing a branch, look at how many people are left on each side — see Anonymity and minimum N.

The demographic axes —sex, age band, tenure band— are the other half of the analysis and do not live in the tree: they cross it. Which means you do not need a «women under 30» node; that is a segment, and you build it while reading the results.

Reorganizing without losing history

Companies reorganize, and the product accounts for it: nodes can be renamed, reordered and moved. Before you delete one, the product tells you what goes with it — how many employees, how many assignments and how many studies touch that branch.

What you cannot do is move the ground under a study that is currently measuring. A study in progress has its scope and its instrument fixed from the moment it launched, and that immobility is what makes its results mean anything — the same reason you cannot edit a question halfway through.