Skip to content

Organising and Sharing Data

Well-organised data is easier to understand, reuse, and share. A clear structure also reduces the risk of files being misplaced, overwritten, or shared with the wrong people.

Plan how your data will be organised and who will need access early in the project, especially when several people or workflows depend on the same files.

Organising Project Data

Keep different types of project files separate. A simple directory structure could look like this:

project-name/
├── README.md
├── data/
│   ├── raw/
│   └── processed/
├── code/
├── models/
├── results/
└── temporary/

The exact structure will depend on your project, but it should clearly distinguish between:

  • original or raw data
  • processed and derived data
  • code and configuration files
  • models and checkpoints
  • results
  • temporary or reproducible files

Treat original data as read-only where possible. Store cleaned, converted, annotated, or otherwise modified files separately.

Keep Temporary Files Separate

Caches, intermediate outputs, extracted archives, and other reproducible files should not be mixed with important project data. Keeping them separate makes them easier to identify and remove later.

Making Data Understandable

Use descriptive and consistent names for files and directories.

Avoid names such as:

final.csv
final_new.csv
final_new_v2.csv
test2/
misc/

Prefer names that explain the contents:

participant_metadata_2026-07-27.csv
images_preprocessed_v2/
resnet50_run03/

Where possible:

  • avoid spaces and special characters
  • use hyphens or underscores consistently
  • write dates in YYYY-MM-DD format
  • include useful version or processing information
  • choose names that are understandable to other project members.

Deciding Who Needs Access

The appropriate storage location and sharing method depend on who needs to work with the data.

flowchart TD
    A{Who needs access to the data?}

    A -->|Only you| B[User storage]
    A -->|Your research group or project| C[Group or project storage]
    A -->|Several ML Cloud users| D[Shared or jointly managed storage]
    A -->|External collaborators| E[Check whether external sharing is permitted]

    E --> F{Are there licence, contractual,<br/>or data protection restrictions?}
    F -->|Yes| G[Clarify the requirements before sharing]
    F -->|No| H[Use an approved sharing or transfer method]

Collaborative data should normally be stored in a group or project location rather than in an individual user directory.

Consider whether everyone needs access to the complete dataset. A smaller subset or read-only access may be sufficient.

For centrally managed datasets used by multiple users or groups, see Shared Datasets.

Choosing the Right Permissions

To give users access to your files, use Access Control Lists.

Give users only the access they need for their work.

Permission Use when
Read A user needs to view or use the data but should not modify it
Write A user needs to add or modify project files
Manage A user is responsible for permissions, structure, or deletion

Before granting access, consider:

  • whether the user needs access to all files
  • whether write access is necessary
  • who may rename or delete data
  • who is responsible for managing permissions
  • whether licence or data protection restrictions apply
  • when the access should be reviewed or removed
Sharing data outside the ML Cloud

Before sending data to an external collaborator or uploading it to another service, check:

  • whether the dataset licence permits sharing
  • whether collaboration agreements restrict access
  • whether the data contains personal, confidential, or medical information
  • whether consent or ethical approval permits the transfer
  • whether the selected service is approved for the data
  • whether encryption or another safeguard is required
  • who will remove access when it is no longer needed

Access to a dataset does not automatically mean that it may be redistributed.

Best Practices

When organising and sharing research data:

  • Use a clear and consistent directory structure
  • Keep original, processed, and result data separate
  • Use descriptive filenames
  • Document the data while the project is active
  • Store collaborative files in group or project storage
  • Grant only the access that is needed
  • Review permissions when project membership changes
  • Check licence and data protection restrictions before sharing
  • Avoid unnecessary duplicate copies

Need Help Organising or Sharing Data?

Contact us through the ticketing system. Include the cluster, storage location, users or groups involved, required access level, and any relevant licence or data restrictions.

Last update: July 27, 2026
Created: July 27, 2026