Skip to content
VIP Assignment Help
Dissertation10 min readUpdated 5 January 2027

How to choose a dissertation topic you can actually finish

Short answer

Dissertation topics

A dissertation topic must pass four tests: the question is narrow enough to answer in your word count, the data is genuinely accessible to you, the method is one you can actually execute, and enough literature exists to build an argument on. Most failed dissertations fail on data access, and they fail in month four — long after the topic was approved and too late to change cheaply.

The four-part feasibility test

Run every candidate topic through these before you commit. It takes an hour and it is the highest-value hour of the whole project.

  • Scope — can you answer this in your word count? A 12,000-word dissertation answers one narrow question well, not three broad ones badly.
  • Data — can you actually get it? Name the specific dataset, the specific participants, or the specific archive. 'I will interview senior managers' is not access; a named organisation that has agreed is.
  • Method — can you execute it with the skills and time you have? A regression you have never run and a transcription workload of forty hours are both real constraints.
  • Literature — is there enough to argue with? Too little and you cannot situate the work; a genuinely untouched topic is usually untouched for a reason.

Data access is where dissertations die

The most common failure pattern: a topic is approved in October, ethics is submitted in January, recruitment starts in February, three people respond, and by April there is no dissertation.

Before you commit, answer this in writing: exactly who or what supplies my data, have I confirmed it, and what is my fallback if they withdraw? A topic without a fallback is a topic with a single point of failure.

Secondary data is not a lesser choice. Large public datasets — national statistics, published surveys, financial filings, existing corpora — remove the recruitment risk entirely and let you spend your time on analysis, which is what is actually marked.

Narrowing a broad interest

Interests start too big. Narrowing is mechanical once you know the moves: add a population, a context, a timeframe, or an outcome measure.

'Social media and mental health' → add a population: 'among first-year undergraduates' → add a context: 'at a UK post-1992 university' → add an outcome: 'and its relationship with reported loneliness' → add a timeframe: 'during the first semester'.

That final version is researchable. The original was a field.

What a good research question looks like

  • It is a question, with a question mark. If you cannot phrase it as one, it is a topic.
  • It is answerable with evidence, not opinion. 'Should universities do X?' is not a research question; 'What effect did X have on Y?' is.
  • It contains its own variables. A reader should see what you are measuring and what against.
  • It is not already definitively answered. Check that first — a supervisor will.
  • It matters to somebody. Name who, in one sentence. If you cannot, the significance section will be painful.

Warning signs to catch early

  • You cannot say in one sentence what you are trying to find out.
  • Your data depends on an organisation that has not agreed in writing.
  • The method requires a skill you plan to learn later.
  • Your literature search returns either three papers or three thousand.
  • The topic is a personal cause you already know the answer to. Confirmation bias is visible to markers, and it caps grades.
  • It requires ethical approval for a vulnerable group, at undergraduate level, on a nine-month timeline.

Choosing with your supervisor rather than at them

Bring three candidate topics, each with a sentence on scope, data, method and literature. That conversation is productive in a way that 'I was thinking about doing something on X' is not.

Ask directly which is most feasible and why. Supervisors have watched dozens of these fail and can usually name the failure mode of a topic in under a minute — but only if you give them something specific enough to assess.

Good and weak versions, side by side

Weak: 'The impact of Brexit on UK business.' Too broad, no population, no measure, no timeframe.

Better: 'How have SMEs in the UK food-import sector adapted their supplier strategies since 2021, and what explains the variation between them?'

Weak: 'Mental health in nursing.' A field, not a question.

Better: 'What factors do newly qualified nurses identify as protective against burnout during their first year of practice?'

In both cases the improvement is the same: a defined population, a defined outcome, a defined boundary.

Frequently asked questions

How do I choose a dissertation topic?
Test each candidate against four criteria: is the question narrow enough for the word count, is the data genuinely accessible to you, can you execute the method, and is there enough literature to argue with. Most failures are data-access failures, and they surface too late to fix.
How narrow should a dissertation question be?
Narrow enough that you could answer it properly in your word count. Add a population, a context, a timeframe and an outcome measure to a broad interest and it usually becomes researchable in four moves.
Can I change my dissertation topic later?
Usually yes, early, and at a cost. Changing after ethics approval or after data collection begins is expensive in time you will not get back — which is why the feasibility test belongs at the start.
Is secondary data analysis a weaker dissertation?
No. It removes recruitment risk and lets you spend your time on analysis, which is what carries the marks. Many of the strongest undergraduate dissertations use existing datasets well.

Keep reading

Related guides