Funder requirements and data protection rules change; confirm current requirements with your research office and your partner's.
Data management plans get written to satisfy a funder and then filed. In a partnered project that is a costly habit, because the plan is what determines who can still use the data in five years.
Write it with your partner, not for them
If you draft the plan alone and send it for sign-off, you will get sign-off. You will not get the local knowledge that makes the plan workable — what storage is realistic at their institution, what their ethics committee requires, what the participants were actually told.
Draft the sections on collection, local storage and consent with the team doing the work, and treat their version as the one to build on.
The sections that matter most
What data will exist, in what formats, at what volume. Listing this properly usually surfaces categories nobody planned for: recordings, photographs, raw instrument output.
Who holds the originals, where, and who has access.
De-identification and who holds the linking key.
Retention and disposal — how long, then what.
Who may use the data for what, during and after the project.
Whether data will be made open, at what level of detail, and where.
Post-project rights are an equity question
This is the section to get right, and the one most often left vague.
Your partner's team generated site access, relationships and local knowledge that cannot be reproduced. Continuing access to the dataset is a large part of what the collaboration is worth to them — it is what their students work on after you have moved to the next grant.
Specify in writing: whether they may publish from the data beyond the joint outputs, whose consent is needed, for how long, and whether they may use it for student training. Write the student training clause explicitly; it is the one people forget and the one with the longest effect.
A default where the foreign institution holds the only working copy and grants access case by case is not a neutral arrangement, even when nobody intends it that way.
Where the data should live
Two designs worth considering, both of which change the default above:
Keep the working dataset on infrastructure in Vietnam, with your team analysing via remote access.
Hold synchronised copies at both institutions, with a clear rule on which is authoritative.
Either arrangement also happens to reduce cross-border transfer problems, since the originals do not leave.
De-identification: check for reidentification risk
Removing names is the first step. Combined fields can still identify individuals — a single unusual case in a small commune, with one distinctive characteristic, is identifiable.
Standard mitigations: aggregate over-detailed fields into bands; suppress very rare categories; and hold the linking key separately with named, limited access and a set deletion date.
Ask your partner to review the reidentification risk. They know the setting and will spot combinations you cannot see from outside.
Open data mandates
If your funder requires open data, three things must be settled before collection, not at publication.
Participant consent must state it. This cannot be retrofitted, and discovering the gap at submission puts your partner in an impossible position with their ethics committee.
The dataset needs documentation — variable definitions, units, missing-value coding. Without it, "open" data is technically published and practically unusable.
For sensitive data, ask your funder about controlled access rather than full openness. Many mandates allow it, and the question is worth asking before you commit your partner to something their consent process cannot support.
Costs belong in the budget
Storage, a data manager's time, transcription, documentation, and the time spent on approvals are all real work. Budget them, including on your partner's side.
Data curation done unpaid gets done late or not at all, and the loss shows up years later when someone tries to reuse the dataset.
A check worth running at the start
Send your partner the plan and ask one question: is there anything here that will not work in practice at your institution?
Ask it privately and make clear you expect to revise. Given the power asymmetry, a plan returned with no comments usually means it was not really reviewed — and the problems will surface during fieldwork instead, when they are expensive.
Who should draft the data management plan?
Draft the collection, local storage and consent sections with the partner team doing the work, and build on their version rather than sending yours for sign-off.
Why are post-project data rights an equity issue?
Continuing access is a large part of what the collaboration is worth to the partner — it is what their students work on after your grant ends.
Is removing names enough to de-identify?
No; combined fields can still identify individuals. Aggregate over-detailed fields, suppress rare categories, and ask the partner to review reidentification risk.
What must be settled before collection if open data is mandated?
Participant consent must state it, the dataset needs proper documentation, and for sensitive data you should ask the funder about controlled access.
