ChatGPT
Perplexity
Claude
Grok
A single real-world project—from the initial question to the final recommendation—will take you further than five completed tutorials. Hiring managers for data analyst roles skim through resumes and look at portfolios. Here’s a structure that stands up to that scrutiny: five steps using a single, end-to-end example, with public data that you can download for free.
As a practical supplement, Tom Gadsby, Senior Data Scientist at CareerFoundry (now part of StackFuel), walks you through, step by step, how to build a compelling data analyst portfolio and what recruiters really look for. The video is in English, but you can turn on subtitles in the player at any time and, if needed, have them automatically translated into German.
We are embedding an external element from a third-party provider here. If you would like to see it, please use the „View content“ button.
What Should Be Included in a Data Analyst's Portfolio?
A compelling data analyst portfolio consists of a single, end-to-end project with five components: a specific question, a clearly documented data cleaning process, a dashboard that draws a conclusion, a one-page recommendation, and a easily accessible location where everything is stored. A single, complete project is better than several half-finished ones.
Step 1: Start with the question, not the dataset
The most common mistake on your first project: You start with the dataset and then look for something interesting. Do it the other way around.
A good portfolio question is specific, relates to real people, and leads to a recommendation. Not „I’ll analyze some housing data,“ but „Where in the city can I get the most square meters for my money, and how has that changed over the years?“
The question should be phrased in such a way that the answer influences a decision. „How many apartments are there?“ is not a good portfolio question. „Where can I get the most space for my budget?“ is a good one, because ultimately, someone can act on that information.
You can find relevant public data for free at GENESIS-Online from the Federal Statistical Office, on the open data portal GovData or among the free data sets at Kaggle. All three are free to use; GENESIS-Online can even be used without registration.
Tip: Choose a topic from an industry you're already familiar with. If you're from the retail sector, use sales data. From logistics? Delivery times. Your industry knowledge makes the analysis more credible because you can tell what's plausible about the numbers and what isn't. That's exactly the advantage you have as a Career changers In the video, Tom offers the same advice from a recruiter's perspective: Choose either a topic you're truly passionate about or an industry that hires a lot of analysts, such as fintech, pharma, e-commerce, or mobility.
Step 2: Refine and document the decisions
Raw data is never clean. And it's exactly that cleanup process that shows you know how to do the job—not the pretty chart at the end.
Open the dataset, and you'll immediately see the typical problems: missing values, prices listed as text instead of numbers, inconsistent spellings for the same district, and a few implausible outliers—such as an apartment for two euros and one for two million.
To get started, the most accessible approach is to use Excel and Power Query; for larger datasets, you’ll perform the same steps in SQL. You remove duplicates, convert text to actual numbers, standardize the district names, and make a conscious decision about what to do with outliers. Don’t just delete them without explanation—justify your decisions.
Important: Document these decisions. One sentence is enough: „Values below ten euros per square meter were removed as input errors.“ Anyone who sees a portfolio in which someone demonstrates how they cleaned up the data and what decisions they made in the process knows that the person has truly understood the job. A perfect dashboard with no visible data processing says very little.
This part isn't just tedious foreplay—it's the heart of the Daily Tasks of a Data Analyst.
Step 3: The Dashboard That Tells a Story
A dashboard isn't an end in itself. It's meant to answer a question. The most common flaw isn't having too little, but having too much: twenty charts, no message.
The rule is: A diagram, a statement. Every visualization must answer a question; otherwise, it’s out. For the question about rent prices, that means: a map or a bar chart showing prices per square meter by district. A line showing the trend over the years. Perhaps a comparison by number of rooms. Nothing more.
And the point that hardly anyone puts into practice: Catchy headlines instead of bland ones. Not „prices by district,“ but „District X offers the most square meters per euro.“ The title already conveys the key insight. This is the pyramid principle: first the statement, then the evidence. Tom refers to exactly this in the video and cites McKinsey’s pyramid principle as a rule of thumb for any summary.
Anyone looking at your dashboard should be able to understand what it says within ten seconds. If they have to study it to figure out what you meant, it’s too complicated. The whole thing is built in Power BI, the tool that handles the bulk of reporting work in everyday business.
Step 4: The recommendation on a single page
This is where a good portfolio stands out from an average one. Most people stop at the dashboard. You go one step further.
Write a summary that is exactly one page long, consisting of four parts: the problem, the key finding, your recommendation, and the next steps.
In our example, this might sound something like this: „If you want to maximize square meters, District X currently offers the best value. Prices there have risen more slowly over the past three years than in the city center. Recommendation for apartment seekers with a budget of Y: check out these three districts first.“
That’s the moment when an analysis becomes a contribution. You’re not just showing that you can analyze data, but that you can use it to make a decision. That’s exactly what data analysts are paid to do.
Tip: Keep it brief. A single page forces you to prioritize the most important points, and that’s exactly what a recruiter wants to see. Someone who clearly summarizes three key insights comes across as more confident than someone who simply lists ten charts side by side without any commentary. And make sure the page is neatly laid out. It’s often the first thing someone reads about your project.
Step 5: Host the site and make it discoverable
Even the best project is useless if it's just sitting on your hard drive. It has to be easy to find and understandable within ten seconds.
You don't need a professionally designed website for this. A simple Notion page, a clean PDF, or the "Featured" section on LinkedIn are more than enough. What's important are clear screenshots of the dashboard, a link to the one-page recommendation, and one sentence per project explaining what question it answers.
Good to know: A single, complete project—one that has been carried through from the initial question to the final recommendation—is better than three half-finished ones. Quality over quantity, always.
Tom's bonus tip from the video: If there’s a company you’d like to work for, ask them directly if there’s a data problem you can work on. Some companies will provide you with anonymized data for this purpose. You’ll be solving a real problem in a field that interests you, and in the best-case scenario, this will be your ticket into the company.
How long does a project like this take?
Realistically, you should plan on spending between 20 and 40 hours on your first complete portfolio project, spread out over two to four weeks. The bulk of that time won’t be spent on the dashboard, but rather on refining the project and really narrowing down the focus. This is normal and mirrors the ratio you’d see in everyday professional life.
Don't let that discourage you. You’re not just creating an application document here—you’re practicing the exact process you’ll need to use every day later on. And you can expand the project at any time: a second data set, a year-over-year comparison, or an additional dimension can be added later without having to start from scratch.
Good to Know: A subsidized continuing education program doesn't take this step out of your hands—it helps you structure it. The portfolio project is an integral part of the curriculum, including feedback on your decisions. This is usually funded through the education voucher, but you're not entitled to it.
An Overview of the Most Common Mistakes
| Error | Why It's Harmful | Better |
|---|---|---|
| Start with the dataset | The result is an analysis without a purpose | Start with a question that leads to a decision |
| Hide Adjustment | That's exactly how you can tell what true craftsmanship is | Document decisions, one sentence each |
| Create twenty charts | Lots of space, no message | A chart, a statement, a call to action |
| Stop at the dashboard | The „So what?“ is missing | One-page recommendation with next steps |
| Save Project Locally | No one thinks so | Notion, PDF, or LinkedIn Featured |
Conclusion: A real project beats five tutorials
Start with the question, not the data. Organize the data and show the decisions you made along the way. Build a dashboard where every chart conveys a message. Summarize the recommendation on a single page. And put it all where someone can grasp it in ten seconds.
Anyone who goes through this entire process from start to finish will have more to show for it than most applicants with a long list of courses.
It's exactly these kinds of projects—using real data sets and incorporating feedback—that you'll build in the portfolio project for the Data Analyst Training. If you want to know if this is right for you, just schedule your free consultation. Together, we'll look at your background, the next step, and the support available through the education voucher. You are not entitled to this; the agency decides on a case-by-case basis. StackFuel has been AZAV-certified since 2020 and has supported more than 8,000 graduates.
Frequently asked questions
How many projects should a data analyst's portfolio include?
One is enough, as long as it's complete. A project carried through from start to finish—from the initial question, through data cleaning and the dashboard, to the recommendation—is more meaningful than three tutorials that were only started.
Where can I get free data sets for my portfolio?
Public sources such as GENESIS-Online from the Federal Statistical Office, the GovData portal, or the open datasets on Kaggle. If possible, choose a topic from an industry you're already familiar with.
Do I need Python for my first portfolio project?
No. Excel with Power Query, SQL, and Power BI are more than enough for a solid first project. Python is an extension, not a starting point.
What do recruiters look at first in a portfolio?
Usually, it's the one-page summary and the dashboard screenshots. That's why both should be neatly laid out and make the key insights immediately apparent.
