A passion project is only worth having if you can be questioned on it for ten minutes without repeating yourself. That rules out most of what gets recommended — the charity founded and abandoned, the club with no members, the app with no users. Twenty-four here that survive the test, grouped by what you would actually produce.
The phrase "passion project" has been hollowed out by people selling them. What survives is a simple test: can you be questioned about it for ten minutes without repeating yourself?
Almost everything in the usual lists fails that.
The test
Four questions, asked in this order by anyone who knows what they are doing.
- What did you actually do?
- What did you find?
- What went wrong?
- What would you do differently?
A real project has long answers to all four. The third is the one that catches people out, because a project that never encountered a problem was not difficult enough to be worth discussing — and a project you did not really do has no answer at all.
Apply the test before you start, not after. If you cannot imagine what question three would be about, pick something else.
The four that never work
The nonprofit you founded. Registered, website built, a launch post, then nothing. The questions — how much did you raise, who did it reach, what happened in year two — have no good answers. Founders of real organisations can answer them; founders of application-shaped ones cannot.
The club with no members. An activity that exists in a list and not in a room.
The app with no users. Building a working app is a genuine achievement. It is also not a finding. "How many people used it, and what did you learn from how they used it?" is the question, and "none" is the usual answer.
The awareness campaign. Almost impossible to evidence. What changed, and how do you know?
None of these are bad things to do. They are bad things to do for an application, because the questioning exposes the gap between the claim and the substance.
Twenty-four that hold up
Grouped by what you would actually produce.
Projects that produce a dataset
These convert into research papers most easily, and they survive questioning because there are numbers to discuss.
- Log every bus on one route against its timetable for three months, and quantify where the delay accumulates.
- Photograph the same stretch of hedgerow weekly for a year and measure what changes and when.
- Record the price of a fixed basket of goods across five local shops monthly, and track divergence.
- Measure air quality outside your school at fixed times for a term, against the nearest official monitor.
- Transcribe and code a year of your local council's planning decisions by outcome and applicant type.
- Collect every job advert in one sector for two months and code what employers actually ask for.
- Track the readability of one newspaper's front page across a decade using archive issues.
- Measure how long it takes to get a GP appointment across a sample of practices, by phoning as a researcher and recording the stated wait.
Projects that analyse data somebody else collected
Faster to start, and the analysis is the contribution.
- Test whether school start times correlate with attendance in districts that changed them.
- Check whether Benford's Law holds across three published financial datasets and explain the departures.
- Compare what two national galleries' wall texts attribute to the artist versus the context.
- Map how coverage of one policy shifted across four newspapers in the six months around its introduction.
- Analyse edit histories on contested Wikipedia articles and characterise who wins.
- Test whether open-source projects with codes of conduct retain contributors differently, using the GitHub API.
- Examine whether constituencies with younger populations voted differently once education is controlled for.
- Compare stated caffeine content on drink labels against published independent measurements.
Our list of free datasets covers where most of this lives.
Projects that make something, with a measurement attached
The making is the hook; the measurement is what makes it discussable.
- Build a weather station and compare six months of your readings against the official station, then explain the differences.
- Write software that does one small useful thing for a real group of people, and report what they did with it.
- Compose, record and release music under a constraint you set, and analyse what the constraint did to the output.
- Translate a short text nobody has translated and write up the decisions you had to make.
- Restore an object and document every decision, including the ones you got wrong.
- Design and run a small teaching intervention in your own school, with a pre- and post-measure.
Projects that are mostly reading, done properly
Underrated, cheap, and strong if the synthesis is genuine.
- Read everything published on one narrow contested question in the last five years and write the honest state of the disagreement.
- Trace how one widely-repeated statistic entered circulation and whether the original source supports it.
That last one is a genuinely excellent project and almost nobody does it. Nearly every field has a number everyone cites and few have checked.
Which group suits which goal
| Group | What you end up with | Converts to a paper? | Cost to start |
|---|---|---|---|
| Produce a dataset | Numbers nobody else has | Yes, directly | Time only |
| Analyse existing data | A finding from public sources | Yes, directly | Nothing |
| Make something, measured | An object plus a measurement | Sometimes | Varies |
| Read properly | A synthesis of a live disagreement | Only at venues taking reviews | Nothing |
What makes any of them work
Three things, independent of which you pick.
A specific output. A dataset, a document, a thing. "I learned about X" is not an output. An interviewer can ask to see an output.
Enough time to go wrong. Three to four months. The questions that reveal a real project are about failures, and a weekend has none.
A question you would want answered anyway. If the only reason to do it is an application, the questioning will find that out — not because interviewers are cynical, but because genuine curiosity produces details that invention does not.
How much does it actually count?
Less than the market implies, and the mechanism is not what students assume.
Caltech reports that 42% of admitted students submitted research and specifies it should come from "a laboratory in a university or research-affiliated organization" with a letter from the research site; independent work is routed to a portfolio instead. Yale asks applicants to state "your specific contributions if the work was done in collaboration with others."
Beyond those, almost no university publishes a position. What does hold up is the interview: a tutor can question you about your project for as long as they want, and a project you genuinely did keeps giving while a borrowed one runs out in ninety seconds. That is the mechanism — not a box being ticked.
For UK applicants specifically, our piece on whether research programmes are worth it in the UK covers what is and is not on the record.
Turning one into a paper
The dataset projects convert; the making projects usually do not.
If publication interests you, pick from the first two groups above, then read the venue's scope before you write. NHSJS and the Columbia Junior Science Journal accept literature reviews — which makes projects 23 and 24 viable — while most student journals want original findings only. The complete guide to publishing sets out what each takes.
If your project needs a laboratory, most universities bar under-eighteens from the relevant work. Research without a lab covers the alternatives, and they are better than waiting.
Where we fit
The Oxford Centre for Advanced Research pairs students with PhD candidates and early-career researchers at Oxford, Cambridge, other leading UK universities and Ivy League institutions. The Oxford Scholar Programme is £2,000 — about $2,600 — for ten contact hours over ten to fourteen weeks.
Nothing on the list above requires us. The twenty-four are deliberately things a student can start this week with no budget, and the dataset ones cost nothing but attention. What a mentor adds is the person who asks question three before you have invested a term — and a teacher with a doctorate will do that free.
Before you pick one
Write down the project, then write the answer you would give to "what went wrong?"
If you cannot imagine an answer, the project is too easy. If you can, you have already started thinking like someone who did it.
Frequently asked questions
What makes a good passion project?
Producing something specific that you can be questioned about. The test is ten minutes of questions — what you did, what you found, what went wrong, what you would do differently. Anything you can exhaust in two minutes is an activity rather than a project, however impressive it sounds in a list.
Does it have to relate to what I want to study?
For UK applications, generally yes — universities select on the subject, so a project connected to it does far more than an unconnected achievement. For US applications there is more room for a genuine outside interest, provided it is real and sustained rather than assembled for the application.
How long should it take?
Long enough to have gone wrong at least once. Three to four months of part-time work is typical. A weekend project has no story in it, and the questions that reveal a real project are the ones about what failed and what you changed in response.
Do I need to found something?
No, and founding things is the most over-recommended and least convincing option. A nonprofit with no accounts, a club with four members or an app with no users invites exactly the questions you cannot answer. Depth in something small beats the appearance of scale.
Can a passion project become a research paper?
Often, if there is data in it. The ones that convert well are the measuring and analysing projects rather than the making ones. If publication interests you, choose a project with a dataset at its centre and the conversion is straightforward rather than a rewrite.
What if I abandon it halfway?
That is fine, and often more interesting than finishing, provided you can say what you learned and why you stopped. An honest account of a project you stopped for good reasons reads better than a project you pretended to finish. What does not survive questioning is a project you quietly dropped and still list.