- Reading time: 3 minutes
- Price: Free download
- Published: 17th September 2026
- Word count: 847 words
- File format: Text
Why do you want to study this course or subject?
The thing that pulled me towards data science was realising how much of the reasoning happens before any model is trained. In my Further Maths statistics work I met the idea that a confidence interval is a statement about a procedure rather than a single answer, and it changed how I read numbers in the news. When a headline says a system predicts something with ninety per cent accuracy, I now want to know what the base rate was, what the errors cost and who was in the training data. Artificial intelligence interests me for the same reason: it is not only a set of techniques but a set of choices about what counts as a good prediction. Reading Hannah Fry's Hello World gave me a clearer sense of how those choices land on real people, particularly in her discussion of how algorithms used in decision-making can be trusted either too much or too little. I want a degree that teaches me the linear algebra, probability and programming properly, rather than only how to call library functions, because I have already found that when my own code produces a nonsense result I can usually only diagnose it if I understand the method underneath. I am also drawn to how broad the field is. The same ideas about representing uncertainty appear in climate modelling, which I met in Geography, and in the recommendation systems I use daily. Studying data science and AI at degree level would let me work on the mathematics seriously while keeping hold of the applied questions that made me interested in the first place.
How have your qualifications and studies helped you to prepare?
My A-levels give me a useful combination. Maths and Further Maths provide the calculus, matrices and statistics that underpin machine learning; working through hypothesis testing and the normal distribution has been the most directly relevant part, and Further Maths introduced me to eigenvalues, which I later recognised in an online explanation of principal component analysis. Computer Science has taught me Python and some Java, along with algorithmic complexity, and my non-exam assessment involved building a revision tool that stored question performance in an SQL database and prioritised topics the user answered poorly. Designing that made me think about sparse data: with only a few attempts per topic the ordering was unstable, so I added a minimum-attempts rule rather than pretending the estimate was reliable. Geography has been unexpectedly valuable. It has trained me in handling real datasets that are incomplete and collected for other purposes, and in writing about evidence with appropriate caution. For my independent investigation I gathered river channel measurements and had to explain honestly why my sample size limited the conclusions I could draw. Outside lessons I have worked through Andrew Ng's introductory machine learning material and used the pandas documentation to teach myself data cleaning, which is slower and less glamorous than modelling but seems to be most of the actual work. I have also started reading about overfitting and cross-validation properly, having first encountered them as words I was using without really understanding.
What else have you done to prepare outside of education, and why are these experiences useful?
My independent project began after several late buses. Using open timetable and vehicle location data published by the regional transport authority, I collected arrival records for three routes near my college over two months and analysed them in Python. I found the average delay was less useful than the spread, because occasional very late buses mattered more to passengers than the typical journey, so I compared medians and the ninetieth percentile by time of day. Merging the datasets was the hardest part: stop identifiers did not always match and I had to decide how to treat missing records rather than silently dropping them. I presented the charts to my Geography teacher, who sensibly asked whether my two-month window included roadworks, which I could not rule out. I now work on the tills at a garden centre most Saturdays and help with the quarterly stock count. It sounds unrelated, but it has taught me a lot about where data comes from. Our stock figures are only as good as the person scanning, and I have seen how a mislabelled tray of bedding plants produces numbers that look precise and are simply wrong. Serving customers has also improved how I explain things: older regulars ask me to work out planting quantities, and I have learned to give a clear answer without a lecture. At college I help run a coding club for Year 12 students, where I set short exercises and debug alongside them. Explaining loops and lists to someone who finds them genuinely confusing has forced me to understand them better myself. I also play in a local badminton league, which keeps my week from being entirely screens and spreadsheets.
This example has 4,736 characters across the three answers. Use it for ideas and structure. Your own UCAS answers must fit within 4,000 characters in total, including spaces.
Why this example works — strengths and ways to improve
This is a thoughtful, well-evidenced statement whose strongest quality is its consistent attention to uncertainty, data quality and method. Each answer stands on distinct evidence and the reflection feels genuine rather than claimed. The main problem is length: at 4,736 characters it exceeds the 4,000-character UCAS limit, so you need disciplined cutting rather than any new material.
Subject motivation
a confidence interval is a statement about a procedure rather than a single answer
You ground your interest in a precise statistical idea and show how it changed your reading of accuracy claims through base rates, error costs and training data. That is real curiosity. The Geography and recommendation-system links are lighter and could be trimmed without weakening the motivation.
Academic preparation
Further Maths introduced me to eigenvalues, which I later recognised in an online explanation
You connect named A-level content to machine learning concepts without overstating your mastery. The NEA and the Geography investigation show applied skill. Your wider study through Andrew Ng's material, pandas and cross-validation is credible, and the admission that you once used terms without understanding them reads honestly.
Evidence and reflection
I added a minimum-attempts rule rather than pretending the estimate was reliable
This is real reflection. You identify a specific weakness in sparse data, make a design decision and explain your reasoning. The bus project does the same with medians and percentiles. These moments show judgement about method, which is far more persuasive than general claims about problem-solving.
Relevant experience
a mislabelled tray of bedding plants produces numbers that look precise and are simply wrong
You turn an ordinary job into strong evidence about data provenance, and the bus project uses public data with sensible caution. The coding club shows that you can explain ideas to others. The badminton sentence adds little and is an easy cut when you reduce the length.
Credibility and voice
slower and less glamorous than modelling but seems to be most of the actual work
Your voice is modest, specific and fits your stage. You never claim professional authority. Your teacher's roadworks question, which you could not answer, strengthens your credibility. The tone stays curious rather than performative throughout.
Structure and format
The three answers are distinct and each comfortably clears 350 characters. However, the 4,736-character total exceeds the 4,000-character UCAS limit by roughly 740 characters. It also sits within the library's 3,500-character minimum, so you can cut about 740 characters and still satisfy both requirements.
What you’ve done well
- You return to uncertainty and data quality in your coursework, your project and your job, which gives the statement coherence without forcing links.
- You describe concrete decisions, such as the minimum-attempts rule, using percentiles and handling missing records, so the reader sees your reasoning in action.
How this draft could improve
- Cut about 740 characters to fit the 4,000-character limit. Strong candidates are the badminton line, the climate and recommendation-system sentence, and some of the scene-setting in the opening.
- In the first answer, condense the Hello World sentence to the single point you took from the book, so that motivation is shown more economically.
- Shorten your description of the coding club activities to one sentence, keeping only your reflection on understanding loops better by explaining them.