Last updated: 2026-10-06

U
Undergraduate level

Finding a Project Idea and Refining the Question

Most projects that go wrong early go wrong at the question, not the code. A project can have a clear plan, a capable student, and good tools, and still answer nothing because the question it started from was vague, too large, or already answered. The work of choosing and refining the question comes first, and it is worth more time than it usually gets.

This page covers three steps: where ideas come from, how to choose one that is worth pursuing, and how to turn a broad topic into a question that can be answered. The next page, Scope, Feasibility, and Defining Success, picks up from there.Scope, Feasibility, and Defining Success

Where Ideas Come From

Good project ideas usually come from one of four places. You may recognise a problem from your own experience, such as a process at work or a tool that frustrates you. A module may have left a question open. A paper may stop short of something you can see how to test. Or a dataset, system, or organisation may be available to you in a way that others are not.

The source matters less than the test that follows. An idea from a personal frustration is only useful if you can state what would change for someone else if it were solved. An idea from a paper is only useful if the paper's loose end is still loose. For the formal side of choosing an overall project, see Choosing a Final-Year Project.see Choosing a Final-Year Project

Selecting a Topic

A topic is a region of interest, such as "recommender systems" or "accessibility in mobile apps". It is not yet a project. A workable topic has three properties:

  • It has an object. There is a specific thing to study or build, not a general field.
  • It has a measure. You can say how you would know whether the thing works, improves, or differs.
  • It has access. You can get the data, the users, the system, or the material you need within the time available.
A quick test. Try to finish this sentence: "I will find out whether ___ changes ___ for ___." If you cannot fill the blanks with something specific, the topic is not yet a project.

Keep two or three candidate topics for a few days. Ask a supervisor or peer which one they would expect to be most informative in either outcome. A project that answers "yes" or "no" with equal value is usually stronger than one that can only succeed.see-also

Refining Broad Topics Into a Question

Once a topic is chosen, it becomes a question by narrowing three things:

  1. The population or system. Which users, which data, which version of the system, which context?
  2. The outcome. What exactly is being measured, and how?
  3. The standard for an answer. What result would count as support for one answer, and what would count against it?

Take "social media and wellbeing" as an example. It is a topic, not a question. Narrowed, it might become a question about whether a particular feature of one platform changes the self-reported mood of a defined group of users over a stated period. The narrowed version is harder to write, but it tells you what data to collect and what the result would mean.too narrow and you hit a dead end

Before committing to a question, check whether it has already been answered. This is the main job of the literature review, covered in Literature Reviews That Do Analytical Work. A question that has been answered well is still a useful question, but the project then has to offer something new, such as a different population, method, or context.

Don't do this. Settle on a question that asks for an opinion ("is X good?") without saying what evidence would settle it. Opinion questions can be written up well, but they need a stated standard for judgement, or the conclusion will read as assertion.

Checking the Question Against Reality

A refined question should survive three checks before you plan around it:

  • Can it be answered in the time available? If answering it needs a year of data collection and you have three months, the question is too big, whatever its merit.
  • Does the answer change something? If every possible result leads to the same action or the same conclusion, the question may not be doing much work.
  • Do you understand the terms? Words such as "engagement", "quality", or "effective" need a definition you can measure before the project starts.

Expect the question to shift once you start reading and building. That is normal. The discipline is to record each change and its reason, rather than letting the question drift without a record. The habit is described in Project Navigation and the Art of Triangulation.

When You Are Stuck

If none of the candidate topics feels right, the problem is usually not a lack of ideas. It is often that the ideas have not been tested against the three properties above. Pick the least bad candidate, write its question as clearly as you can, and test it for one week with a small piece of reading or a rough prototype. The week will tell you more than another evening of deliberation. For a fuller treatment of stalling and switching tracks, see Getting Unstuck and The Unstuck Ladder.