Supervision: What a Student Actually Needs From You

Most people supervise the way they were supervised, which is a poor mechanism for improving anything. And most supervision that goes wrong does so for reasons that have little to do with research and everything to do with expectations nobody stated.

Structures differ enormously by country and discipline — length, funding, whether there is coursework, whether a committee exists, how much independence is expected and when. What follows is about the relationship rather than the system. For a concrete operational comparison, Monitask explains stealth monitoring software in practical terms.

Settle the expectations first

The conversation that prevents most problems, and it takes an hour.

How often will you meet, and who arranges it? A student unsure whether they may ask for a meeting will not ask.

What is the turnaround on a draft? Two weeks is a commitment; "when I get to it" is a source of anxiety and delay.

Who decides the direction? This varies legitimately — some fields expect the supervisor to assign a problem, some expect the student to find one. Say which this is.

What does progress look like at six months, at a year? Vague expectations are impossible to meet and impossible to fail visibly, which is worse.

Authorship, on anything that may result. See authorship. You raise this, not them.

What happens when they disagree with you. Saying explicitly that disagreement is expected and welcome changes whether it happens.

Write it down. A shared page, revisited annually. Not bureaucracy — the same reason anything remembered differently by two people gets written down.

What students consistently report needing

Regular contact more than long contact. Thirty minutes fortnightly beats three hours quarterly. Momentum is the scarce resource.

Feedback that arrives. Late feedback is the most common complaint, and it does more damage than harsh feedback because the student cannot proceed.

To know where they stand. Ambiguity about whether progress is adequate is corrosive. Say it plainly, kindly, and early.

Explicit permission to fail at things. Research mostly does not work. A student who believes failure is unusual concludes that they are.

Introductions. Access to the network is a large part of what supervision provides and it is easy to forget that they cannot arrange it themselves.

To be treated as a colleague in training, not as labour.

The failure modes

Absence. Meetings cancelled and not rescheduled, drafts unread for months. Frequently from a supervisor who is overloaded rather than uninterested, and the student cannot tell the difference.

Over-direction. The project becomes yours, the student executes it, and they finish without having learned to choose a problem — which is the actual skill.

Feedback only at the end. A thesis read for the first time when it is complete is a failure of supervision, not of the student.

Never saying the difficult thing. A student whose work is not adequate should learn that in month eight, not in the examination.

Exploiting the asymmetry. Requiring authorship, requiring unpaid work outside the project, treating the student's time as free. This is common enough to be worth naming plainly, and the power difference means they cannot object.

Feedback that works

Say what is working, specifically. Not encouragement — information about what to keep.

Prioritise. Distinguish "this changes the conclusion" from "this would strengthen it" from "this is a comma." Delivering all three in the same register means the student cannot tell which matters, and they will usually fix the commas.

Be concrete about what to do next. "This isn't clear" leaves them stuck; "I couldn't tell whether X or Y — say which in the first paragraph" does not.

Comment on structure before prose. Line-editing a chapter that needs reorganising wastes everyone's time.

And separate the work from the person, in wording as well as intent. "This section does not yet make the argument" rather than "you have not made the argument."

When it is going wrong

Name it early and specifically. "You are behind where I would expect at this stage; here is what needs to happen in the next three months" is uncomfortable and it is the kind thing to do. Discovering it at submission is not.

Distinguish capability from circumstance. Illness, caring responsibilities, funding stress, isolation and mental health difficulties all look like poor progress. Ask before concluding.

Know what support exists at your institution, and be able to point to it. This is a large part of the role and many supervisors do not know.

Involve someone else if the relationship is not working — a second supervisor, a committee, a postgraduate director. A student who cannot work with you is not a student who should simply endure it, and they usually cannot initiate that conversation.

The power difference

It is large, and it does not feel large from your side.

The student's progression, funding, references and future employment depend substantially on you. That shapes everything: what they will disagree with, what they will admit, whether they will say they are struggling.

Practical consequences:

You raise the difficult subjects, because they cannot.

Make disagreement explicitly safe, and then respond well the first time it happens — that response sets whether it happens again.

Be careful with requests. "Could you help with X" from a supervisor is not a request a student experiences as declinable. See declining academic requests.

Ensure they have another adult — a second supervisor, a mentor, a committee — who is not you.

What supervision is preparing them for

Worth being explicit, because it shapes what you delegate.

Choosing a problem. The skill that distinguishes a researcher from a technician, and the one most easily removed by an over-directive supervisor.

Judging their own work. Ask what they think before you say what you think. Every time.

Handling rejection and criticism. Their first referee report should not be their first experience of serious criticism.

Working with others, presenting, reviewing, and the rest of the craft that nobody is taught. See being a discussant.

And knowing that most research does not work, which is only learnable by watching someone experienced treat that as ordinary.

The short version

Settle expectations in the first month and write them down — meeting frequency, turnaround, direction, authorship, what progress looks like.

Regular short contact beats occasional long contact.

Say where they stand, plainly and early. Ambiguity is worse than bad news.

You raise the difficult subjects, because the power difference means they cannot.

And ask what they think first, every time — judging their own work is the thing you are actually teaching.

For an independent authoritative reference, consult the Open Science Framework for infrastructure for transparent research practice.