A three-year collaboration ends. There are four papers, a final report, and a set of results. The funder is satisfied.
A separate question: what can the partner institution do now that it could not do before?
For many projects the honest answer is: very little. That outcome is not accidental, and it is decided at proposal stage rather than at the end.
Why capacity is rarely designed in
Four reasons:
Funders evaluate outputs. Papers, datasets, patents. Capacity left behind is harder to measure, so it goes unmeasured and therefore unplanned.
It costs time in the short run. Doing an analysis yourself is faster than teaching someone to do it — and there is a deadline.
Nobody's job. The principal investigator is responsible for deliverables; nobody is responsible for what the partner is left with.
It is assumed to happen automatically. The belief that working alongside someone transfers capability is widespread and mostly wrong — proximity is not training.
What actually persists
Six things, roughly in order of how durable they are:
People who learned to do something. The most durable, and the most portable. A technician who can run a method independently keeps that whether or not the equipment survives.
Methods embedded in local practice — written protocols, adapted to what is actually available locally, in the working language of the people who will use them.
A local person who has led something. Someone who has been principal investigator on a work package, chaired a session, written a first draft — this changes what they can do next, and what they can apply for.
Published work with the partner in a leading position. A first-authored paper is credential that opens doors after you are gone.
Connections beyond you. If the partner's only international link is your group, the collaboration ending means isolation resuming.
Equipment — the least durable, for reasons below.
The equipment problem
Equipment is the most visible form of legacy and the one that most often fails. Four things determine whether it survives:
Consumables. An instrument requiring reagents the partner cannot afford or cannot import becomes furniture within a year.
Maintenance and service. Is there anyone in the country who can service it? A machine that fails and cannot be repaired locally is a total loss.
Power and environment. Instruments requiring stable power or specific conditions fail where those are not guaranteed.
Someone trained and still present. If the one person who knew how to run it has moved on, the capability leaves with them — so train more than one.
Practical rule: ask the partner what would actually be useful before buying anything. The answer is frequently something less impressive and more useful than what you had in mind, and they know which of their constraints will kill which machine.
Budgeting for it
Five lines that make capacity a real plan rather than an aspiration:
Staff time for training — on both sides. Teaching takes hours that have to exist in the budget.
Exchange visits in both directions. Note the second direction: sending your people there builds less than bringing their people here, but doing only the first is the pattern that produces no transfer.
Publication costs for partner-led papers.
Conference travel for partner staff, including junior members — this is where connections beyond you get built.
Consumables and maintenance beyond the project period, where the funder permits.
Training that works
Four principles:
Train more than one person. Single points of failure are how capacity evaporates.
Have them do it, not watch it. Including the parts that go wrong — troubleshooting is most of the skill.
Write it down locally. A protocol written by the person who will use it, in their language, referencing the equipment they actually have.
Follow up after you leave. A scheduled call three months later, when the first real problems appear, is worth more than another week of on-site training.
Who leads what
This is where the difference between a genuine partnership and subcontracting becomes visible.
Three things to plan deliberately:
At least one work package led by the partner — with real decision authority, not nominal leadership.
At least one paper with a partner first author, planned from the start rather than offered at the end.
The partner presenting the work at an international meeting, at least once.
None of these costs anything extra. All three change what the partner can do afterwards.
How to tell whether anything was left
Two years after the grant ends, ask:
Has the partner run anything using what the project built — without us?
A new grant of their own, a paper we are not on, a method still in use, a student trained on it. If the answer is no on all four, the collaboration produced outputs and left nothing behind.
That is a legitimate way to run a project — but it should be a choice made openly, not a default arrived at by never asking the question.
Why is capacity rarely designed into collaborations?
Because funders evaluate outputs, training costs time against a deadline, nobody is formally responsible for it, and people assume proximity transfers capability — which it does not.
What form of legacy is most durable?
People who learned to do something — a technician who can run a method independently keeps that whether or not the equipment survives.
What determines whether donated equipment survives?
Affordable consumables, locally available servicing, suitable power and environment, and more than one trained person still present.
How do you tell whether anything was left behind?
Two years after the grant ends, ask whether the partner has run anything using what the project built without you — a grant, a paper you are not on, a method still in use.
