Data Story From Your Previous Field Using Public Data

The problem career changers have with data portfolios

Entry-level data roles are competitive, and many portfolios contain the same few datasets analysed the same way: Titanic survival, a superstore sales file, a movie ratings table. Recruiters have seen them hundreds of times. Career changers have an advantage they rarely use: they already know a field. A former accountant knows which expense ratios are misleading, a teacher knows which attendance patterns predict dropping out, a pharmacist knows which medicine shortages recur, and a logistics coordinator knows where delivery times really go.

A project that asks a question only someone from that field would think of, and answers it in that field's language, stands out immediately. It also answers the interview question "why data, and why you" before it is asked.

What already exists and the gap

• Our World in Data publishes cleaned datasets and charts on health, education, energy and more.
• Kaggle Datasets hosts thousands of public datasets and notebooks, many of them repeated analyses of the same files.
• Tableau Public has a gallery of published dashboards you can learn from.
• Google Data Analytics Professional Certificate ends with a case study, but most learners pick the same suggested scenarios.

The raw material is abundant. What is missing is domain judgement: knowing which question matters and which number misleads. That is what you bring, and what this project shows.

What you build

1. The question: one sentence, plus who would act on the answer. "Which districts' schools lose the most students between grades 5 and 6, and is it linked to distance to the nearest middle school?" is a good question. "Explore education data" is not.
2. The data: a public source with a clear licence: a national statistics bureau, a government open data portal, an international organisation or a research repository. Write down what the data does not tell you: missing years, changed definitions, self-reported figures.
3. Cleaning: load into SQL or pandas and keep a log of every change and why.
4. Analysis: answer the question, then try to prove yourself wrong with a second cut of the data.
5. Charts and summary: three to five charts labelled for a busy reader, and a one-page summary with the answer, how confident you are and what would make you more confident.

The technically interesting part

The interesting part is the check that could have proved you wrong. Most beginner analyses stop at the first chart that supports the story. Yours should include a deliberate test: a different year, a different region, a control group, or a confounder you know from your field. If the pattern survives, you can say so with more confidence. If it does not, you have found something more interesting, and you have shown the habit that separates analysts people trust from analysts who make pretty charts.

How to evaluate it

• Reproducibility: someone else can clone the repository and get the same numbers in under 15 minutes.
• Reader test: give the one-page summary to 3 people who are not analysts, ideally from your old field. Can they state the finding and its main caveat in one sentence each?
• Honesty: every limitation of the data is written down, and the summary's confidence matches the evidence.
• Timing: a reviewer can understand the README in five minutes.

Suggested plan

1. Week 1: write the question, find and document the dataset.
2. Week 2: clean, log changes, first answer.
3. Week 3: the disproving check, charts, first draft of the summary.
4. Week 4: reader test, revise, publish repository and dashboard.

To stretch this into a final year project, repeat it for three related questions, add a published dashboard and a data refresh script, and run a proper reader study.

Path to a product

This is first a portfolio piece. It can lead to freelance work: NGOs, trade bodies, newsrooms and consultancies in your old field often need someone who understands both the domain and the data, and will pay per report. A realistic first customer is your former employer or a professional association you belonged to. A regular data newsletter for your field is another route.

Scope warning

Do not build a machine learning model, a web app or a scraper. Do not use private data from a former employer, even anonymised, without written permission. One question, answered well, beats ten shallow charts.

Learn what you need on Sefism

• Data analytics roadmap : the SQL, spreadsheet and visualisation skills in order.
• Data analytics cheat sheet : quick reference while you clean and query.
• Storytelling with Data : how to make charts a busy manager understands.
• Our World in Data : documented public datasets to start from.
• Can I get into tech without a computer science degree : context for career changers building a first portfolio.

Get help with this idea

The hardest step is choosing the question. Tauseef can help you turn your field experience into a sharp question, find usable data and plan an evaluation that shows real judgement. If you are new here, create a free account first.

• Discuss this idea with a mentor
• Send a query to Tauseef
• Book a 1:1 session