Hiring managers for analyst roles want proof of three things: you can get messy data into shape, you can find something worth knowing in it, and you can explain it to someone who does not read SQL. A portfolio is where you show the third, and the third is where most candidates fall short.
Pick projects that answer a business question
Start each project from a question a manager would ask, then find data for it. "Which product categories lose money after returns?" makes a better project than "Exploratory analysis of a retail dataset." Three good sources of questions:
- Public datasets from government portals, such as data.gov.in, or a city's open data site.
- Your own work, with company data replaced by a public or synthetic stand-in.
- A small business you know: a shop's sales register, a club's membership list, a college fest's ticket sales.
Write each project in five parts
- The question, in one sentence.
- The data: where it came from, how many rows, and what you had to clean.
- The method: the tools and the steps, kept short. SQL, Python, Excel, Power BI or Tableau, named where you used them.
- The finding, with a number: "Electronics returns wipe out 60% of the category's margin."
- The recommendation: what you would tell the business to do next.
Put the finding near the top of each project. A hiring manager skimming three projects should be able to read three findings without scrolling.
Show the work behind it
Link the notebook or SQL file on GitHub, and a live dashboard if you built one. Add one chart image on the page itself, because many readers will not click through. Label the axes and give the chart a title that states the finding.
What to leave out
- The Titanic, Iris and other tutorial datasets, unless you asked a question nobody else did.
- Ten dashboards with no write-up. A dashboard without a finding is a screenshot.
- Tool logos in place of results. "Python, Pandas, NumPy, Matplotlib" belongs under the method, once.
Put it online today
Your portfolio needs a summary, your experience, the projects and contact links. If your resume already has those, Dossier turns it into a page in a few minutes, and you can edit each project's text to the five-part format above. For the keywords analyst job posts repeat, see the data analyst resume keyword list.
Questions people ask
- What should be in a data analyst portfolio?
- Two to four projects, each with a business question, the data source, the method, a finding with a number and a recommendation. Add links to the code or dashboard, plus a short summary, experience and contact details.
- Do I need a website for a data analyst portfolio, or is GitHub enough?
- GitHub holds the code, and hiring managers rarely read notebooks first. A short web page with the findings up front gets read; link GitHub from it for anyone who wants the detail.
- Where can I find data for portfolio projects?
- Government open data portals, city data sites, public company filings and anonymised data from your own work or a small business you know. Choose data that lets you answer a real business question.