How to find research collaborators outside your institution

Most research collaborations are not advertised; they are inherited. A supervisor suggests someone, a colleague down the corridor joins in, a name from a conference comes back a year later. That works, but it caps your pool at the people you already know. The moment a study needs a second site, an expertise your department does not have, or a question that crosses into another field, the inherited network runs out. This is about what happens after that point.

Define the role before you look for the person

"I am looking for a good collaborator" is not something you can search for. What is the concrete deficit in the study: someone to compute the sample size, a second site to collect data, a specialist in a method you do not use, or someone who already holds the dataset? Naming the missing role narrows the candidate list on its own, and it makes the first email specific instead of vague.

Name the contribution at the same time. The CRediT taxonomy^[1] describes contributions such as conceptualization, data curation, formal analysis, software and writing across 14 roles. CRediT labels contributions; it does not decide authorship. Authorship criteria are defined separately, and in biomedical fields the ICMJE recommendations are widely used^[2]. The value of naming roles early is that the authorship conversation happens at the start, when it is still a planning question rather than a dispute.

Where candidates actually are

The most productive source is the set of papers you already read. Look at the author lists of work published in your area in the last two or three years. A recent paper is a strong signal that the person still cares about the topic, though it is worth checking their latest output and any ongoing projects as well, since publication can trail the work by years. Corresponding author addresses appear on most papers; confirm the address is still current on the institutional page before writing.

The second source is open bibliographic databases. Subject searches in OpenAlex^[3] or Google Scholar show who is productive and where they sit. The thing to watch here is that you want productivity in your topic, not overall, which I come back to below.

The third source is society working groups and the early-career networks that most conferences run. Poster sessions beat talks for this: you get five minutes with the person and, at a poster, that person is usually the one doing the actual work.

There is also a layer people forget: one step beyond your own network. Your supervisor's former students, your co-authors' co-authors. Writing to them with "you may know my name through our shared work with X" gets a substantially better response rate than a fully cold approach.

How to assess a candidate

The common mistake is ranking candidates by total citations or h-index. Those measure a whole career; they do not measure what someone can contribute to your study. Narrower and more useful signals:

• Publications in your specific topic. How many pieces of work has this person produced on your question? A giant of the field who has never touched your problem is less useful than a mid-career researcher with five papers on it.
• Activity in the last two or three years. People who left a topic five years ago rarely come back. Recent output means current interest.
• Shared language and time zone. Trivial on paper, decisive in practice. They set the pace of correspondence and revision rounds.
• Seniority balance. Mid-career researchers who are active on the topic can often contribute directly and tend to be more open to new collaborations. If you approach a very senior name, identify the team member who will actually run the work as well.
• What the institution adds. Case volume, equipment, a cohort, an archive: the infrastructure behind the person is part of the team.

A first email that gets a reply

The only job of the first email is to let the reader answer "is this a serious proposal" in thirty seconds. Five to seven sentences is enough. One sentence on who you are; one on why you are writing to them specifically, ideally citing one of their papers; two or three on the proposal itself (role, estimated workload, what you are offering); one clear question at the end. Do not attach the protocol or a long CV. They will ask if they want them.

Dear Dr ___, I read your paper on ___ in ___. I work on ___ at ___, and we are planning to test ___ with a two-site design. The contribution we anticipate from your site is ___, with an estimated workload of ___. We propose defining contributions with CRediT roles from the start and settling authorship on the basis of the contributions that actually happen. Would you like to see a short protocol outline?

Silence is not personal; everyone's inbox is full. Send one polite reminder after two weeks. If there is still no answer, move to the next candidate rather than pressing.

Red flags

Some signals tell you early that a collaboration will not reach the manuscript stage:

• Someone who asks for authorship without describing a contribution, or says "just put my name on it". Publication ethics bodies treat this under honorary authorship^[4].
• Reluctance to discuss data ownership and author order. If that conversation does not happen at the start of a project, it happens at the end as an argument.
• "You go ahead, get ethics approval, we will sort it out later." That is an absence of commitment.
• Disappearing for weeks during the very first exchange. It does not get faster once the study is running.

Sources and standards

1. CRediT (Contributor Roles Taxonomy), NISO — defines contributions across 14 roles. credit.niso.org
2. ICMJE, "Defining the Role of Authors and Contributors" — authorship criteria. icmje.org
3. OpenAlex — open catalogue of scholarly works, authors, institutions and topics. openalex.org
4. COPE (Committee on Publication Ethics) — guidance on authorship disputes. publicationethics.org
5. ORCID — persistent researcher identifier and publication record. orcid.org