Data analyst portfolio: which projects to build and how to present them
Hiring managers do not need another notebook that reproduces a tutorial. They want to see you take a real question, a messy dataset and a deadline, and come back with a recommendation.
The short answer: a data analyst portfolio is three or four projects, each starting from a question someone would actually pay to have answered, showing how you found and cleaned the data, what you analysed, what you found and what you would recommend. Between them they should show the tools the jobs you want ask for, usually SQL, a spreadsheet, a programming language such as Python or R and a visualisation tool. Put the code and notebooks on GitHub, and tie it all together on a short portfolio site that a hiring manager can read in a few minutes without opening a single notebook.
What a hiring manager is looking for
An analyst's job is not to produce charts. It is to help someone make a better decision. So a reviewer reads your portfolio for four things:
- Framing: can you turn a vague problem into a question that data can answer?
- Handling real data: missing values, duplicates, inconsistent categories, joins that do not quite line up. Clean tutorial data hides all of this.
- Sound analysis: the right method for the question, and awareness of what the data cannot tell you.
- Communication: a finding and a recommendation a non-analyst can follow, in a few sentences and one or two clear charts.
Choosing the projects
Three or four strong projects beat ten exercises. Choose them so that, together, they cover the skills in the postings you are applying to, and so that each one is about a subject you can talk about with some interest in an interview.
| Project type | What it shows | Example of a question |
|---|---|---|
| A SQL analysis | Joins, aggregation, window functions, querying a database rather than a single file | Which customer groups stop ordering after their first month, and when? |
| A dashboard | Choosing measures, designing for a reader, building something people would use weekly | How are a city's bike-share journeys changing by station and season? |
| A cleaning and analysis project in Python or R | Wrangling messy data, statistics, reproducible work | Did a change in a public service's opening hours affect how many people used it? |
| A business case | Turning numbers into a recommendation with costs and risks | Should a small shop extend its weekend hours, based on its sales data? |
The examples are illustrations, not a list to copy. The datasets that every course uses make it hard to stand out, because a reviewer has seen the same analysis many times. If you use one, ask a question of it that nobody else asks. Better still, find data on a topic you know: your local council's open data, a sport you follow, a hobby, a public dataset from a field you want to work in.
How to write up each project
- The questionOne or two sentences on what you set out to answer and who would care about the answer.
- The dataWhere it came from, how big it is, the period it covers and its limitations.
- What you didThe cleaning that mattered, the approach and why you chose it. Mention one thing that went wrong and how you handled it.
- What you foundTwo or three findings in plain language, each with the chart that shows it best.
- What you would recommendThe decision the findings support, and what you would want to check next.
- The linksThe repository, the notebook, the live dashboard. Each one should work for someone who is not signed in.
Put the findings first. A hiring manager should get the answer from the top of the page, and the method from further down if they want it.
GitHub or a portfolio website?
Both, doing different jobs. GitHub holds the evidence: the SQL, the notebooks, the cleaning scripts. GitHub's own documentation describes a README as the place to tell people why a project is useful, what it does and how to get started with it (GitHub Docs, checked on 2026-11-04), so give every portfolio repository one that leads with the question and the findings.
A portfolio site is the summary a recruiter or hiring manager actually reads. Many of them will not clone a repository or scroll through a notebook, but they will read a page with your name, what kind of analyst you are, three projects with their findings and charts, your skills and a resume download. Link each project on the site to its repository, and the repository back to the site.
The rest of the page
- A headline that says what kind of analyst you are, such as "Data analyst focused on retail and customer behaviour".
- Skills grouped as a reviewer scans them: querying, analysis, visualisation, tools. Skip percentage bars.
- Experience, including non-analyst roles where you used data to make a decision.
- Writing, if you publish short analyses or explainers. It shows communication more clearly than anything else.
- A resume download and one contact.
Common mistakes
- Projects with no question, only a dataset and some charts.
- Notebooks without a summary, which ask the reader to run your code to find out what you found.
- Dashboards with every possible chart rather than the three that answer the question.
- Claiming certainty the data does not support. Say what the limitations are.
- Dead links to dashboards or notebooks that require signing in.
- Publishing data you did not have the right to share.
Building it on Introfio
Each project fits a Work entry on Introfio: The problem for the question, What I did for the approach, The result for findings and recommendation, up to 8 Key numbers, and up to 8 Screenshots of your charts, each with a caption. Live link can point at a published dashboard and Source code at the repository, and Tools & technology lists SQL, Python or whatever you used. A Skills section groups your tools, and a Resume section offers your resume as a download.
The Developer and Portfolio templates are both free and suit analysts: Developer sets your projects on a changelog timeline, Portfolio leads with your name set very large and an index of numbered work. If you write short analyses, the site's own blog supports headings, images and code blocks, with 3 published posts on Free and unlimited on Pro. The developers page and the job seekers page show the rest of the set-up, and the developer portfolio guide goes deeper on writing up technical projects.
Common questions
How many projects should a data analyst portfolio have?
Three or four strong projects are usually enough, chosen so that together they cover the tools in the jobs you want, such as SQL, a spreadsheet, Python or R and a dashboard tool.
Is GitHub enough for a data analyst portfolio?
GitHub holds the code, but many reviewers will not open a notebook. A short portfolio site with each project's question, findings and charts, linking to the repositories, is easier for them to read.
Where can I find data for portfolio projects?
Public open-data portals, datasets from a field you want to work in, or data on a topic you know well. Check the licence before publishing analysis, and never use an employer's or client's data.
Do I need a portfolio if I already work as an analyst?
It matters less once you have relevant experience, but a few written-up projects still help when you change industry or move into a more senior role, as long as you never publish an employer's data.
Keep reading
- How to build a developer portfolio website that gets readWhat hiring managers look for in a developer portfolio, how to choose projects, how to write each entry with stack, role and links, and the mistakes to avoid.
- Resume vs portfolio vs CV: what each one is, and when you need itA resume summarises, a CV records, a portfolio proves. What each one contains, how long it should be, which a job asks for, and how to make them work together.
- How to put your portfolio link on a resume, and in every applicationWhere a portfolio link goes on a resume, how to write it so it works on paper and on screen, what to say in the email, and how to check it before you apply.
Build your site on Introfio
Pick a template, fill it in, and publish on your own link. Free to start, no card needed.