Before the download, write a data note.

A dataset arrives with a context

A file can be easy to download and still be difficult to use responsibly. A useful first step is a short data note stored with the project. Describe the source, the version you obtained, the question it was collected to address and the conditions attached to its use. Record where the authoritative information lives instead of relying on a copied sentence in a chat.

Keep the note small enough that the team will maintain it. A page of clear decisions is more useful than an elaborate document nobody reads. It should help a collaborator understand what the files are, why they are present and what they may do with them.

Separate originals from working copies

Make the source material distinguishable from cleaned files, derived measurements and figures. Avoid silently editing a raw file to make an analysis run. If a correction is necessary, preserve a record of the change and the reason for it. An unexplained difference between two copies can waste more time than a modest amount of organisation at the beginning.

Use identifiers consistently across files. Decide where a mapping table belongs and who needs access to it. A convenient folder structure is not itself a privacy control. When the work involves information about people, use the institution’s approved process rather than improvising a sharing arrangement.

Plan the handover while the project is small

List what you expect another person to need: input access, derived data, code, a description of the environment and the decisions behind exclusions. These items may require different sharing arrangements. Publishing code does not mean the accompanying participant data should be public too.

For a teaching exercise, consider whether synthetic or appropriately released demonstration data can serve the purpose. State the limitations of that substitution. If you cannot share an input, explain how an authorised researcher could request access and what can still be inspected without it.

Review what actually leaves the folder

Before a release, inspect the package rather than assuming that the intended contents are the actual contents. Notes, filenames and logs can contain information that was not meant for publication. Ask a second person to review the handover when the material is sensitive or the workflow is unfamiliar.

NIH’s data-management guidance is one example of a funder describing sharing expectations and limits. It is not a universal rulebook for every Swiss project. The relevant institution, funder, consent conditions and agreements determine what applies. The practical habit is broader: decide early, document the decision and revisit it when the project changes.