How to build a data analyst portfolio, with UK project ideas

A data analyst portfolio is a small set of projects, usually on GitHub, that shows you can take a question, find and clean real data, analyse it and explain what you found. Aim for three finished projects on questions you chose yourself, using the tools employers ask for, such as SQL, Excel, Power BI and Python. The UK publishes plenty of free data to use, from ONS statistics to house prices and electricity data, and each project needs a clear README that says what you asked, what you did and what you found.

What should a data analyst portfolio include?#

Newcastle University’s careers service recommends building a portfolio of work to show your skills and interest, for example on GitHub. For a junior analyst, the portfolio should show the everyday work of the job. The civil service’s data analyst role profile lists managing, cleaning and aggregating data, using code, checking that data is fit for purpose, and presenting it clearly with visualisations.

Across your projects, try to show:

• SQL, in at least one project, because it’s how analysts get data out of databases.
• A dashboard in Power BI or Tableau, built for a named kind of user.
• Cleaning, with the steps you took and why.
• Excel or Python for analysis, depending on the jobs you want.
• Clear writing: findings with numbers, a recommendation and the limits of the data.

Each project should start from a question someone might really ask, such as “which parts of the city centre are busiest on Saturday afternoons?”, rather than from a tool or a dataset. If you’re still deciding whether this is the job for you, start with our guide on how to become a graduate data analyst.

How many projects do you need?#

Three good projects are a sensible target, and quality matters far more than quantity. A finished project with a clear answer beats five half-built ones. GitHub lets you pin up to six repositories to the top of your profile, so put your best three there.

A good set of three might be:

1. A SQL project that joins and summarises several tables.
2. A dashboard project with a short written summary for a non-technical reader.
3. A project in the sector you want to work in, or one that collects data automatically, if you’re interested in more technical roles.

Five project ideas using free UK data#

All five are free to download. Check the licence for anything you use and credit the source in your README. HM Land Registry and Leeds City Council publish their data here under the Open Government Licence, which lets you copy, publish and adapt the information, including commercially, as long as you acknowledge the source. The National Energy System Operator (NESO) licenses its Carbon Intensity API under Creative Commons Attribution 4.0, which also asks you to credit it.

1. How has online shopping changed since 2019?

The Office for National Statistics publishes a monthly time series of internet sales as a percentage of total retail sales in Great Britain, which you can download as a spreadsheet. Chart the trend, look for seasonal patterns and describe how the share has changed since 2019. Add a second ONS series, such as total retail sales, to put the numbers in context. This suits Excel or Python, and a one-page written summary.

2. When and where do road collisions happen?

The Department for Transport’s road safety data on data.gov.uk, now called the National Data Library, covers personal injury collisions on public roads in Great Britain that were reported to the police, back to 1979. It comes as separate collision, vehicle and casualty files that use numeric codes, with a data guide to decode them. Load a few years into a database, join the three tables in SQL and compare collisions in your area by hour, weekday and type of road. Keep the tone careful: every row is a real incident.

3. Where have house prices risen fastest?

HM Land Registry’s Price Paid Data records residential property sales in England and Wales back to January 1995, with the price, date, postcode, property type, whether it was a new build and whether it was freehold or leasehold. Start with a yearly file rather than the full download, which is over 5 GB, and note that the files come without column headers by default: HM Land Registry’s guide to the Price Paid Data lists them. Compare median prices by area over ten years, or new builds against existing homes. Use medians rather than averages, because a few very expensive sales can pull an average up.

4. When is electricity greenest?

The Carbon Intensity API from NESO, which took over from National Grid ESO in October 2024, gives forecasts and estimates of the carbon intensity of electricity in 14 regions of Great Britain, for each half hour. Collect a few weeks of data with Python, then show which times of day and which regions are greenest, and how the forecasts compare with later estimates. It’s a good first step towards data engineering, because you have to collect and store the data yourself.

5. How busy is your city centre?

Local councils publish open data too. For example, Leeds City Council publishes some of its data on Data Mill North, including hourly footfall counts from eight cameras around the city centre, in monthly files. Combine the files, then find the busiest hours and days, compare weekdays with weekends and track the trend. The publisher notes that some cameras had recording problems and were moved, and that it has revised the weekly data back to 2020, which is exactly the kind of limit to explain in your README. Your own council may publish similar data, so search its website.

How do you present a project on GitHub?#

Put the effort into the README. GitHub’s guide to READMEs says a README is often the first thing a visitor sees in your repository, so write its first few lines for someone who may read no further.

A README that works for most analyst projects:

1. A title that says what you found or asked, such as “Leeds city centre footfall: when is it busiest?”
2. The question and who would care, in two or three sentences.
3. The data: its name, who publishes it, a link, the date you downloaded it and its licence. If a publisher doesn’t give its own wording, the Open Government Licence says to use “Contains public sector information licensed under the Open Government Licence v3.0.”
4. What you did: cleaning, analysis and tools, in a few bullet points.
5. What you found: two or three findings with numbers, and a chart or dashboard screenshot.
6. Limits and next steps: what the data can’t tell you, and what you would do with more time.
7. How to run it: the files, what to install and in what order to run things.

Keep the repository tidy: clear file names, no large raw data files (link to the source instead), and notebooks with headings, explanations between the code cells and their outputs saved, so charts show without anyone running them.

Then set up your profile. GitHub shows a profile README at the top of your profile when you create a public repository with the same name as your username and add a README.md file. Use it for two or three lines about you and links to your best projects.

Which dashboards should you publish?#

Publish one or two dashboards that answer a clear question, not every chart you’ve made.

• Tableau. Tableau Public is free, and everything you publish there is public, so it only suits open data.
• Power BI. Power BI Desktop is free to download, but sharing online through the Power BI service needs a work or school account: Microsoft says you can’t sign up with a personal email address such as Gmail or Outlook.com. Put screenshots in your README and include the .pbix file, so anyone with Power BI Desktop can open it.

If you do have access to the Power BI service, remember that Publish to web makes a report viewable by anyone on the internet, and Microsoft warns not to publish confidential or proprietary information that way.

Whichever tool you use, give every chart a title that states the finding, label the axes and units, and don’t rely on colour alone to show a difference.

What makes a portfolio hard to read?#

• No README, or one that only says “My project”.
• Notebooks with no explanation, or without saved outputs, so nothing shows until someone runs them.
• Only practice datasets that appear in many courses.
• Clutter: unfinished repositories, copies of course exercises and forks at the top of your profile.
• Missing sources or licences, or data you weren’t allowed to share.
• Charts without titles, labels or units.
• Findings with no numbers, or conclusions the data can’t support, such as claiming one thing caused another from a correlation.

How can you check your portfolio?#

When your projects are up, check your GitHub portfolio with the free Portfolio Check. You type in your GitHub username, and it scores your public projects the way a hiring manager for junior data roles skims them, then suggests the three fixes that matter most.

Then put your best projects on your CV. Our guide on how to write a data analyst CV shows how to list them, with a full example.