01Interest is not the binding constraint.
Students choose topics by interest, which is reasonable and insufficient, because interest sustains a project for about three months and a capstone runs considerably longer than that. What sustains it afterwards is being able to make progress, and progress depends on access rather than enthusiasm.
The question that decides most capstones is therefore not what do I care about. It is what can I actually obtain: which people will talk to me, which data exists and is reachable, whose permission is required and how long it takes to get.
Somebody who asks that in week two chooses a topic they can finish. Somebody who asks it in month five discovers that the interesting question requires access they will not get, and has to choose again with a year gone.
02Three tests before committing.
Each answerable in an afternoon, and each has ended more bad projects than any supervisor's scepticism.
- The access test. Name the specific people or records the study needs, and the specific person who controls access to them. If you cannot name that person, you do not have a project yet.
- The permission test. Establish what approval is required and roughly how long it takes, including any organisational permission on top of the board's. That timeline sits on your critical path and it is not shortenable by working harder.
- The refusal test. Ask what happens if the site says no, and have an answer. A capstone with one possible setting is a capstone with a single point of failure, which is the same arithmetic that governs a placement.
- And a fourth worth adding: could this be answered with what already exists. Some perfectly good capstones require no new collection at all, and they finish.
03Narrow beats broad, and by more than people expect.
The commonest fault in a proposed topic is width. A question about how something affects outcomes across a sector is not a capstone; it is a research programme, and the version a student submits is inevitably thin because the question was never answerable at that scale.
A narrow question feels less impressive and produces a far better piece of work, because it can actually be answered with the evidence available to one person in one year. Examiners are not looking for scope. They are looking for a question, a method capable of answering it, and a conclusion that follows.
There is a diagnostic for this while searching: if every round of reading produces entirely new names with no overlap, the topic is too wide, and the answer is to narrow rather than to read more. That signal is described in the citation chaining issue.
04The four pieces that have to agree.
The problem, the question, the method and whatever you propose doing about it. At the outset they usually do agree, because they were conceived together in one sitting.
What happens afterwards is drift. A supervisor suggests tightening the problem and you tighten it. Some weeks on, a second reader wants the question made more precise and you oblige. Every one of those requests is reasonable, nobody goes back to check the remaining three against the change, and after a few rounds the four are describing separate studies.
That is the single commonest reason a capstone stops, and it is why the four should be written on one page and re-read as a paragraph after every round of feedback. Two minutes, repeated, prevents the thing that costs a year. Where it has already happened, repairing the alignment is a smaller job than restarting, which is what the capstone desk looks at first.
05Choosing something you can bear to live with.
Having argued that interest is not the binding constraint, it is still a constraint. You will be reading this literature for a long time, explaining it at social occasions, and returning to it on evenings when you would rather not.
So the topic should be one you can tolerate at length rather than one you find thrilling at the outset, and those are different properties. Thrilling topics are frequently thrilling because they are broad, which is exactly the fault above.
The combination worth aiming for is a narrow question inside an area you find genuinely interesting. The area supplies the endurance; the narrowness supplies the finishability. Choosing on interest alone gets the first without the second, which is how most abandoned capstones were selected.
Questions at this desk.
What if my supervisor approved a topic I now think is too wide?
Raise it early rather than working at it. Supervisors approve a great many proposals and generally welcome a student narrowing one, because they have watched wide topics fail before. The conversation is much easier in month two than in month eight, when narrowing means discarding work.
Can I use existing data instead of collecting my own?
Frequently, and it is underused. Secondary analysis, document study and evaluations of existing material all produce legitimate capstones and remove the entire access-and-permission timeline. Where your programme permits it, it is the single most reliable route to finishing on schedule.
How narrow is too narrow?
Narrower than most students fear, and the genuine floor is whether an answer would tell anybody anything. A question with a foreseeable and uninteresting answer is too narrow; a question answerable with your access that somebody in the field would want to know is exactly right.
What if my access falls through halfway?
That is why the refusal test exists at the outset. Where it has happened without a fallback, the options are a second site, a change of method to something using available material, or a narrower version of the same question. All three are survivable and all three are faster if raised immediately.
Notes in the margin, from the people at the desks.
READ THE LETTER