Skip to main content

Label Studio Organization & Workspace Structure Guide

Oussama Assili
Updated

This guide explains how to design your Label Studio Enterprise organization and workspace hierarchy to mirror your company's business units, departments, or projects — while maintaining security, clarity, and scalability.

Label Studio Enterprise supports a flexible organizational model that allows teams to collaborate securely on data labeling and ML training tasks.

1. Key concepts

Concept Description Typical example
Organization The top-level entity that groups users, workspaces, and resources under one umbrella. Each organization is fully isolated from others. Acme Corporation
Workspace A logical container for labeling projects, users, and datasets within an organization. Usually represents a business unit, department, or product line. Marketing, R&D, Quality Assurance
Project A collection of labeling tasks and configurations within a workspace. Represents a specific data labeling initiative. Sentiment Analysis for Campaign Feedback
User role Defines access privileges within an org, workspace, or project. Owner, Administrator, Manager, Reviewer, Annotator

2. Recommended structure: mirroring business units

The goal is to create a structure that reflects your company's organizational model while isolating data access between units.

Example hierarchy

Organization: Acme Corporation
│
├── Workspace: Marketing
│   ├── Project: Customer Sentiment Analysis
│   ├── Project: Social Media Image Tagging
│   └── Members:
│       ├── Manager: marketing_manager@acme.com
│       ├── Annotators: marketing_team@acme.com
│       └── Reviewer: analytics@acme.com
│
├── Workspace: R&D
│   ├── Project: Product Feature Classification
│   ├── Project: Model Evaluation
│   └── Members:
│       ├── Manager: rnd_lead@acme.com
│       ├── Annotators: data_scientists@acme.com
│       └── Reviewer: leadership@acme.com
│
└── Workspace: Quality Assurance
    ├── Project: Annotation Review
    └── Members:
        ├── Manager: qa_lead@acme.com
        ├── Annotators: qa_team@acme.com
        └── Reviewer: compliance@acme.com

This separation ensures data from one unit is not visible to another, access is granularly controlled, and workspaces can be independently managed and audited.

3. Roles and least-privilege access

Label Studio Enterprise has five organization-level roles. For the full permissions matrix, see User roles and permissions.

Role Description Recommended assignment
Owner Manages the organization with full permissions at all levels. Not an assignable role — there is only one Owner per organization, tied to the account that created it. Changing the Owner requires a support ticket. The account that created the organization
Administrator Full permissions at most levels: can access and update all workspaces and projects, invite members, and set most organization settings. Central IT or Platform Engineering
Manager Full administrative access over projects and workspaces they created or were added to. Cannot access the Organization page. Business unit or team lead
Reviewer Reviews annotated tasks; can only view projects with tasks assigned to them. QA or ML Ops teams
Annotator Labels tasks; can only view and label projects with tasks assigned to them. Labeling staff, data annotators

Best practice: Assign users the lowest level of access they need to perform their job. For example, if a user only needs to review completed work, assign them as a Reviewer, not a Manager.

4. Workspace creation workflow

  1. Determine business alignment — identify major business units or departments (e.g., Marketing, R&D, QA) and plan one workspace per unit.
  2. Create workspaces — from the Organization page, create a new workspace, name it after the business unit, and add a concise description.
  3. Add members — add members via email or SSO/SCIM, and assign roles based on function (Manager for the unit lead, Annotator for labeling team members, Reviewer for QA/auditors).
  4. Set project ownership — each workspace can have multiple projects; consider limiting project creation to Managers to maintain structure and avoid duplication.

5. Cross-workspace collaboration

If users need access to multiple workspaces, assign roles per workspace rather than globally. For example, a data scientist could be a Reviewer in Marketing and an Annotator in R&D — this prevents unnecessary data exposure while enabling collaboration where appropriate.

6. Auditing & governance

  • Review membership periodically to remove inactive users.
  • Audit roles regularly to ensure least privilege.
  • Enable activity logging to track project activity.
  • Use naming conventions for clarity, e.g. workspaces as Dept-Name and projects as Workspace_ProjectPurpose_YYYYMM.

Draft for review. Corrected: the original draft referenced a "Viewer" role, which doesn't exist in Label Studio Enterprise. Actual roles are Owner, Administrator, Manager, Reviewer, and Annotator — role table and examples updated to match.

Related to